Ni Hms 479919

Inequality in Landownership, the Emergence of Human-Capital
Promoting Institutions, and the Great Divergence

ODED GALOR,
Brown University
OMER MOAV, and
Hebrew University, Royal Holloway University of London, and Shalem Center
DIETRICH VOLLRATH
University of Houston
Abstract
This paper suggests that inequality in the distribution of landownership adversely affected the
emergence of human-capital promoting institutions (e.g. public schooling), and thus the pace and
the nature of the transition from an agricultural to an industrial economy, contributing to the
emergence of the great divergence in income per capita across countries. The prediction of the
theory regarding the adverse effect of the concentration of landownership on education
expenditure is established empirically based on evidence from the beginning of the 20th century in
the U.S.
1. INTRODUCTION
The last two centuries have been characterized by a great divergence in income per capita
across the globe. The ratio of GDP per capita between the richest and the poorest regions of
the world has widened considerably from a modest 3 to 1 ratio in 1820 to an 18 to 1 ratio in
2001 (Maddison, 2001). The role of geographical and institutional factors, human-capital
formation, ethnic, linguistic, and religious fractionalization, colonization, and globalization
has been the centre of a debate about the origin of the differential timing of the transition
from stagnation to growth and the remarkable change in the world income distribution.
This paper suggests that inequality in the distribution of landownership adversely affected
the emergence of human-capital promoting institutions (public schooling and child labour
regulations), and thus the pace and the nature of the transition from an agricultural to an
industrial economy, contributing to the emergence of the great divergence in income per
capita across countries.
1
The theory further suggests that some land-abundant countries that
were characterized by an unequal distribution of land were overtaken in the process of
industrialization by land-scarce countries in which land distribution was rather equal.
The transition from an agricultural to an industrial economy has changed the nature of the
main economic conflict in society. Unlike the agrarian economy, which was characterized
by a conflict of interest between the landed aristocracy and the masses, the process of
2009 The Review of Economic Studies Limited
1
Most of the existing studies (e.g. Hall and Jones, 1999), attribute the differences in income per capita across countries largely to
differences in total factor productivity (TFP), whereas some (e.g. Manuelli and Seshadri, 2005) provide evidence in favour of the
dominating role of human capital. Nevertheless, it should be noted that even if the direct role of human capital is limited, it has a large
indirect effect on growth via its effect on technological progress and the implementation of growth-enhancing institutions (Glaeser, La
Porta, Lopez-De-Silanes and Shleifer, 2004).
NIH Public Access
Author Manuscript
Rev Econ Stud. Author manuscript; available in PMC 2013 August 12.
Published in final edited form as:
Rev Econ Stud. 2009 January ; 76(1): 143179. doi:10.1111/j.1467-937X.2008.00506.x.
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
industrialization has brought about an additional conflict between the entrenched landed
elite and the emerging capitalist elite. The capitalists who were striving for an educated
labour force supported policies that promoted the education of the masses, whereas
landowners, whose interest lay in the reduction of the mobility of the rural labour force,
favoured policies that deprived the masses from education.
The process of industrialization raised the importance of human capital in the production
process, reflecting its complementarity with physical capital and technology. Investment in
human capital, however, has been suboptimal due to credit market imperfections, and public
investment in education has been therefore growth enhancing.
2
Nevertheless, human-capital
accumulation has not benefited all sectors of the economy. In light of a lower degree of
complementarity between human capital and land,
3
a rise in the level of education increased
the productivity of labour in industrial production more than in agriculture, decreasing the
return to land due to labour migration and the associated rise in wages. Landowners,
therefore, had no economic incentives to support these growth enhancing educational
policies as long as their stake in the productivity of the industrial sector was insufficient.
4
The proposed theory suggests that the adverse effect of the implementation of public
education on landowners income from agricultural production is magnified by the
concentration of landownership.
5
Hence, as long as landowners affect the political process
and thereby the implementation of education reforms, inequality in the distribution of
landownership is a hurdle for human-capital accumulation, slowing the process of
industrialization and the transition to modern growth.
6
Economies in which land was rather equally distributed implemented earlier public
education and benefited from the emergence of a skilled-intensive industrial sector and a
rapid process of development. In contrast, among economies marked by an unequal
distribution of landownership, land abundance that was a source of richness in early stages
of development, led in later stages to under-investment in human capital, an unskilled-
intensive industrial sector, and a slower growth process. Thus, variations in the distribution
of landownership across countries generated variations in the industrial composition of the
economy, and thereby the observed diverging development patterns across the globe.
7
The prediction of the theory regarding the adverse effect of the concentration of
landownership on education expenditure is confirmed empirically based on data from the
beginning of the 20th century in the U.S. Variations in public spending on education across
states in the U.S. during the high-school movement are utilized in order to examine the
thesis that inequality in the distribution of landownership was a hurdle for public investment
in human capital. In addition, historical evidence suggests that, indeed, the distribution of
2
See Galor and Zeira (1993), Fernandez and Rogerson (1996), Benabou (2000), and Galor and Moav (2004).
3
Although rapid technological change in the agricultural sector may increase the return to human capital (e.g. Foster and Rosenzweig,
1996), the return to education is typically lower in the agricultural sector, as evident by the distribution of employment. For instance,
as reported by the U.S. Department of Agriculture (1998), 569% of agricultural employment consists of high-school dropouts, in
contrast to an average of 137% in the economy as a whole. Similarly, 166% of agricultural employment consists of workers with 13
or more years of schooling, in contrast to an average of 545% in the economy as a whole.
4
Landowners may benefit from the economic development of other segments of the economy due to capital ownership, households
labour supply to the industrial sector, the provision of public goods, and demand spillover from economic development of the urban
sector.
5
The proposed mechanism focuses on the emergence of public education. Alternatively, one could have focused on child labour
regulation, linking it to human-capital formation as in Doepke and Zilibotti (2005), or on the endogenous abolishment of slavery (e.g.
Lagerlof, 2003) and the incentives it creates for investment in human capital.
6
Consistent with the proposed theory, Besley and Burgess (2000) find that over the period 19581992 in India, land reforms have
raised agricultural wages, despite an adverse effect on agricultural output.
7
As established by Chanda and Dalgaard (2008), variations in the allocation of inputs between the agriculture and the non-agriculture
sectors are important determinants of international differences in TFP, accounting for between 30% and 50% of these variations.
GALOR et al. Page 2
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
landownership affected the nature of the transition from an agrarian to an industrial
economy and has been significant in the emergence of sustained differences in human-
capital formation and growth patterns across countries.
The next section places the research in the context of the existing literature. Sections 3 and 4
develop the theoretical model and its testable predictions. Section 5 provides anecdotal
historical evidence that is consistent with the proposed hypothesis. Section 6 examines
empirically the hypothesis that the concentration of landownership had an adverse effect on
education expenditure based on the U.S. experience during the high-school movement, and
Section 7 offers some concluding remarks.
2. RELATED LITERATURE
The central role of human-capital formation in the transition from stagnation to growth is
underlined in unified growth theory (Galor, 2005). This research establishes theoretically
(Galor and Weil, 2000; Galor and Moav, 2002) and quantitatively (Doepke, 2004;
Fernandez-Villaverde, 2005; Lagerlof, 2006) that the rise in the demand for human capital in
the process of industrialization and its effect on human-capital formation, technological
progress, and the onset of the demographic transition have been the prime forces in the
transition from stagnation to growth. As the demand for human capital emerged, variations
in the extensiveness of human-capital formation and therefore in the rate of technological
progress and the timing of the demographic transition significantly affected the distribution
of income in the world economy (Voigtlander and Voth, 2006; Galor and Mountford, 2006,
2008).
The proposed theory suggests that the concentration of landownership has been a major
hurdle in the emergence of human-capital promoting institutions. Thus the observed
variations in human-capital formation and in the emergence of divergence and overtaking in
economic performance is attributed to the historical differences in the distribution of
landownership across countries. In addition to our own findings that land inequality had a
significant adverse effect on education expenditure in the U.S., the predictions of the theory
are consistent with the findings by Deininger and Squire (1998) and Easterly (2007) about
the inverse relationship across countries between land inequality (across landowners), on the
one hand, and human-capital formation and growth, on the other hand.
8
The role of institutional factors has been the focus of an alternative hypothesis regarding the
origin of the great divergence. North (1981), Landes (1998), Mokyr (1990, 2002), Parente
and Prescott (2000), Glaeser and Shleifer (2002), and Acemoglu, Johnson and Robinson
(2005) have argued that institutions that facilitated the protection of property rights,
enhancing technological research and the diffusion of knowledge, have been the prime
factor that enabled the earlier European take-off and the great technological divergence
across the globe.
The effect of geographical factors on economic growth and the great divergence have been
emphasized by Jones (1981), Diamond (1997), Sachs and Warner (1995), and Hibbs and
Olsson (2005). The geographical hypothesis suggests that favourable geographical
conditions permitted an earlier transition to agriculture in Europe and made it less vulnerable
to the risk associated with climate and diseases, leading to the early European take-off,
whereas adverse geographical conditions in disadvantageous regions, generated permanent
hurdles for the process of development.
8
Furthermore, Banerjee and Iyer (2005) show that historically landlord-dominated districts of West Bengal in India fare worse on
agricultural productivity and schooling than small-holder districts.
GALOR et al. Page 3
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
The exogenous nature of the geographical factors and the inherent endogeneity of the
institutional factors led researchers to hypothesize that initial geographical conditions had a
persistent effect on the quality of institutions, leading to divergence and overtaking in
economic performance.
9
Engerman and Sokoloff (2000), ES, provide descriptive evidence
that geographical conditions that led to income inequality, brought about oppressive
institutions (e.g. restricted access to the democratic process and to education) designed to
maintain the political power of the elite and to preserve the existing inequality, whereas
geographical characteristics that generated an equal distribution of income led to the
emergence of growth promoting institutions. Acemoglu et al. (2005), AJR, provide evidence
that reversals in economic performance across countries have a colonial origin, reflecting
institutional reversals that were introduced by European colonialism across the globe.
Reversals of fortune reflect the imposition of extractive institutions by the European
colonialists in regions where favourable geographical conditions led to prosperity, and the
implementation of growth enhancing institutions in poorer regions.
10
The proposed theory differs in several important dimensions from the earlier analysis of the
relationship between geographical factors, inequality, and institutions. First, it suggests that
a conflict of interest between landowners and landless individuals, and in particular, among
the economic elites (i.e. industrialists and landowners), rather than between the ruling elite
and the masses as argued by ES and AJR, brought about the delay in the implementation of
growth enhancing educational policies.
11
Hence, in contrast to the viewpoint of ES and AJR
about the persistent desirability of extractive institutions for the ruling elite, the proposed
theory suggests that the implementation of growth-promoting institutions emerges in the
process of development as the economic interest of the two elites in the efficient operation
of the industrial sector dominates. Second, consistent with existing cross-sectional evidence
and the evidence presented in this paper, the theory underlines the adverse effect of unequal
distribution of landownership (rather than wealth inequality as suggested by ES) in the
timing of educational reforms. Third, the theory focuses on the direct economic incentive
(i.e. the adverse effect of education reforms on the land rental rate) that induces the landed
elite to block education reforms, rather than on the effect of political reforms on the
distribution of political power and thus the degree of rent extraction. Hence, unlike ES, and
AJR, even if the political structure remains unchanged, economic development may
ultimately trigger the implementation of growth promoting institutions.
12
A complementary approach suggests that interest groups (e.g. landed aristocracy and
monopolies) block the introduction of new technologies and superior institutions in order to
protect their political power and thus maintain their rent extraction. Olson (1982), Mokyr
(1990), Parente and Prescott (2000), and Acemoglu and Robinson (2006) argue that this type
of conflict, in the context of technology adoption, has played an important role throughout
the evolution of industrial societies.
13
Interestingly, the political economy interpretation of
9
The role of ethnic, linguistic, and religious fractionalization in the emergence of divergence and growth tragedies has been linked
to their effect on the quality of institutions (Easterly and Levine, 1997).
10
Additional aspects of the role of colonialism in comparative developments are analysed by Bertocchi and Canova (2002). Brezis,
Krugman and Tsiddon (1993), in contrast, attribute technological leapfrogging to the acquired comparative advantage of the current
technological leaders in the use of the existing technologies (via learning by doing).
11
The role of a conflict of interest within economic elites in economic and political transformation was examined earlier by Lizzeri
and Persico (2004), Llavador and Oxoby (2005) and others.
12
In contrast to the political economy mechanism proposed by Persson and Tabellini (2000), where land concentration induces
landowners to divert resources in their favour via distortionary taxation, in the proposed theory land concentration induces lower
taxation so as to assure lower public expenditure on education, resulting in a lower economic growth. The proposed theory is therefore
consistent with empirical findings that taxation is positively related to economic growth and negatively to inequality (e.g. Benabou,
1996; Perotti, 1996). Bowles (1978) discusses the incentives of landlords to restrict access to education in order to preserve a
relatively cheap labour force.
13
Barriers to technological adoption that may lead to divergence are explored by Caselli and Coleman (2001), Howitt and Mayer-
Foulkes (2005) and Acemoglu, Aghion and Zilibotti (2006) as well.
GALOR et al. Page 4
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
our theory suggests, in contrast, that the industrial elite would relinquish power to the
masses in order to overcome the desire of the landed elite to block economic development.
14
Empirical research is inconclusive about the significance of human capital rather than
institutional factors in the process of development. Some researchers suggest that initial
geographical conditions affected the current economic performance primarily via their effect
on institutions. Acemoglu et al. (2005), Easterly and Levine (2003), and Rodrik,
Subramanian and Trebbi (2004) provide evidence that variations in the contemporary
growth processes across countries can be attributed to institutional factors whereas
geographical factors are secondary, operating primarily via variations in institutions.
Moreover, Easterly and Levine (1997), and Alesina, Devleeschauwer, Easterly, Kurlat and
Wacziarg (2003) demonstrate that geopolitical factors brought about a high degree of
fractionalization in some regions of the world, leading to the implementation of institutions
that are not conducive for economic growth and thereby to diverging growth paths across
regions.
Glaeser et al. (2004) revisit the debate whether political institutions cause economic growth,
or whether, alternatively, growth and human-capital accumulation lead to institutional
improvement. In contrast to earlier studies, they find that human capital is a more
fundamental source of growth than political institutions (i.e. risk of expropriation by the
government, government effectiveness, and constraints on the executives). Moreover, they
argue that poor countries emerge from poverty through good policies (e.g. human-capital
promoting policies) and only subsequently improve their political institutions.
Finally, the paper contributes to the political economy approach to the relationship between
inequality, redistribution, and economic growth. This literature argued initially that
inequality generates political pressure to adopt redistributive policies and that the
distortionary taxation that is associated with these policies adversely affects investment and
economic growth (Alesina and Rodrik, 1994; Persson and Tabellini, 1994). Existing
evidence, however, does not support either of the two underlying mechanisms (Perotti,
1996). In contrast, the proposed theory suggests that inequality (in the distribution of
landownership) is in fact a barrier for redistribution and growth promoting educational
policy, provided that landowners have sufficient political power. This mechanism resembles
the one advanced by Benabou (2000) in his exploration of the relationship between
redistribution and growth. He demonstrates that a country would implement an efficient tax
policy and converge to a higher income steady state, provided that the initial level of
inequality is low and that the better-endowed agents have therefore limited interest to lobby
against it. Otherwise the efficient redistribution will be blocked, perpetuating initial
inequality and confining the economy to a low-income steady state.
15
3. THE BASIC STRUCTURE OF THE MODEL
Consider an overlapping generations economy in a process of development. In every period
the economy produces a single homogeneous good that can be used for consumption and
investment. The good is produced in an agricultural sector and in a manufacturing sector
using land, physical, and human capital as well as raw labour. The stock of physical capital
in every period is the output produced in the preceding period net of consumption and
14
See Lizzeri and Persico (2004) and Ghosal and Proto (2005) as well.
15
This mechanism is echoed in Gradstein (2007) which argues that the support for the protection of property rights is greater the more
equal is the distribution of income and the smaller is the political bias. Similarly, Bourguignon and Verdier (2000) suggest that if
political participation is determined by the education (socioeconomic status) of citizens, the elite may not find it beneficial to subsidize
universal public education despite the existence of positive externalities from human capital. See also Benabou (2002) for the trade-
offs between redistribution and economic growth.
GALOR et al. Page 5
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
human-capital investment, whereas the stock of human capital in every period is determined
by the aggregate public investment in education in the preceding period. The supply of land
is fixed over time. Physical-capital accumulation raises the demand for human capital and
output grows due to the accumulation of physical and human capital.
16
At the outset, the economy consists of three groups of individuals: a homogeneous group of
landowners, a homogeneous group of landless capitalists and workers who are landless and
do not own capital initially. In the process of development, physical capital is accumulated
by all groups.
3.1. Production of final output
The output in the economy in period t, y
t
, is given by the aggregate output in the agricultural
sector, , and in the manufacturing sector, ,
(1)
3.1.1. The agricultural sectorProduction in the agricultural sector occurs within a
period according to a neoclassical, constant-returns-to-scale (CRS) production technology,
using labour and land as inputs. The output produced at time t, , is
(2)
where X
t
and L
t
are land and the number of workers, respectively, employed by the
agricultural sector in period t. Hence, workers productivity in the agricultural sector is
independent of their level of human capital. The production function is strictly increasing
and concave, the two factors are complements in the production process, F
X L
> 0, and the
function satisfies the neoclassical boundary conditions that assure the existence of an
interior solution to the producers profit-maximization problem.
Producers in the agricultural sector operate in a perfectly competitive environment. Given
the wage rate per worker, , and the rate of return to land,
t
, producers in period t choose
the level of employment of labour, L
t
, and land, X
t
, so as to maximize profits. That is, {X
t
,
L
t
} = argmax[F(X
t
, L
t
) w
t
L
t

t
X
t
]. The producers inverse demand for factors of
production is therefore,
(3)
3.1.2. Manufacturing sectorProduction in the manufacturing sector occurs within a
period according to a neoclassical, CRS, CobbDouglas production technology using
physical and human capital as inputs.
17
The output produced at time t, , is
16
Alternatively, the rise in the demand for human capital could have been based on technological progress, and output growth could
have been due to technological progress and factor accumulation. This specification would not alter the main qualitative results.
17
As will become apparent, the choice of a CobbDouglas production function assures that there is no conflict of interest among
landless individuals regarding the optimal education policy, permitting the analysis to focus on the conflict between the landowners
and the landless.
GALOR et al. Page 6
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
(4)
where K
t
and H
t
are the quantities of physical capital and human capital (measured in
efficiency units) employed in production at time t. Physical capital depreciates fully after
one period. In contrast to the agricultural sector, human capital has a positive effect on
workers productivity in the manufacturing sector.
Producers in the manufacturing sector operate in a perfectly competitive environment. Given
the wage rate per efficiency unit of labour, , and the rate of return to capital, R
t
,
producers in period t choose the level of employment of capital, K
t
, and the number of
efficiency units of labour, H
t
, so as to maximize profits. That is,
. The producers inverse demand for factors of
production is therefore
(5)
3.2. Individuals
In every period a generation, which consists of a continuum of individuals of measure 1, is
born. Individuals live for two periods. Each individual has a single parent and a single child.
Individuals, within as well as across generations, are identical in their preferences and innate
abilities, but they may differ in their wealth.
Preferences of individual i who is born in period t (a member i of generation t) are defined
over second-period consumption, , and a transfer to the offspring, .
18
They are
represented by a log-linear utility function
(6)
where (0, 1).
In the first period of their lives individuals acquire human capital. In the second period of
their lives individuals join the labour force, allocating the resulting wage income, along with
their return to capital and land, between consumption and income transfer to their children.
In addition, individuals transfer their entire stock of land to their offspring.
19
An individual i born in period t receives a transfer, , in the first period of life. A fraction t
t
0 of this capital transfer is collected by the government in order to finance public
18
This form of altruistic bequest motive (i.e. the joy of giving) is the common form in the recent literature on income distribution
and growth. It is supported empirically by Altonji, Hayashi and Kotlikoff (1997). As discussed in Section 4, if individuals generate
utility from the utility of their offspring the qualitative results remain intact. First period consumption may be viewed as part of the
consumption of the parent.
19
This assumption captures the well-established observation (e.g. Bertocchi, 2006) that at least in early stages of development land is
not fully tradable due to agency and moral hazard problems. It is designed to assure that landowners could be meaningfully defined as
a distinct viable class. In the presence of a market for land, the anticipation of education reforms and the associated decline in rental
rates would generate a decline in the price of land. Thus, as long as land is not fully tradable, landowners who would be the prime
losers from the decline in the price of land would object to education reforms. If land would be fully traded, land holdings would be
equivalent to any other asset holdings, and in contrast to historical evidence, landowners would not be a significant force in the
political structure of the economy. The proportion of land-holding in the portfolio of each individual should not vary systematically
across groups, and thus efficient education policy will be implemented.
GALOR et al. Page 7
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
education, whereas a fraction 1 t
t
is saved for future income.
20
Individuals devote their
first period for the acquisition of human capital. Education is provided publicly free of
charge. The acquired level of human capital increases with the real resources invested in
public education. The number of efficiency units of human capital of each member of
generation t in period t + 1, h
t+1
, is a strictly increasing, strictly concave function of the
government real expenditure on education per member of generation t, e
t
.
21
(7)
where h(0) = 1, lim
e
t
0
+ h(e
t
) = , and lim
e
t
h(e
t
) = 0. Hence, even in the absence of
real expenditure on public education individuals still posses one efficiency unit of human
capitalbasic skillsassuring the operation of the industrial sector in every time period.
In the second period of life, members of generation t join the labour force earning the
competitive market wage w
t+1
. In addition, individual i derives income from capital
ownership, , and from the return on landownership, x
i

t+1
, where x
i
is the
quantity of land owned by individual i. The individuals second-period income, , is
therefore
(8)
A member i of generation t allocates second-period income between consumption, , and
transfers to the offspring, , so as to maximize utility subject to the second-period budget
constraint:
(9)
Hence, the optimal transfer of a member i of generation t is,
22
(10)
consumption , and the indirect utility function of a member i of generation t,
, is therefore monotonically increasing in :
(11)
where (1 ) ln(1 ) + ln.
20
As discussed below, an income tax rather than a bequest tax would complicate the analysis, but would not alter the qualitative
results.
21
A more realistic formulation would link the cost of education to (teachers) wages, which may vary in the process of development.
As can be derived from Section 3.4, under both formulations the optimal expenditure on education, e
t
, is an increasing function of the
capitallabour ratio in the economy, and the qualitative results remain therefore intact.
22
Note that individuals preferences defined over the transfer to the offspring, , or over net transfer, , are represented in
an indistinguishable manner by the log-linear utility function. Under both definitions of preferences the bequest function is given by
.
GALOR et al. Page 8
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
3.3. Physical capital, human capital, and output
The aggregate level of intergenerational transfers in period t, as follows from (10), is a
fraction of the aggregate level of income y
t
. A fraction
t
of this capital transfer is
collected by the government in order to finance public education, whereas a fraction 1
t
is
saved for future consumption. The capital stock in period t + 1, K
t+1
, is therefore
(12)
whereas the government tax revenues are
t
y
t
.
Let
t+1
be the fraction (and the numbersince population is normalized to 1) of workers
employed in the manufacturing sector. The education expenditure per young individual in
period t, et, is,
(13)
and the stock of human capital, employed in the manufacturing sector in period t + 1, H
t+1
,
is therefore,
(14)
Hence, output in the manufacturing sector in period t + 1 is,
(15)
and the physicalhuman capital ratio k
t+1
K
t+1
/H
t+1
is,
(16)
where k
t+1
is strictly decreasing in
t
and in
t+1
, and strictly increasing in y
t
. As follows
from (5), the capital share in the manufacturing sector is
(17)
and the labour share in the manufacturing sector is given by
(18)
The supply of labour to agriculture, L
t+1
, is equal to 1
t+1
, and the supply of land is fixed
over time at a level X > 0. Output in the agriculture sector in period t + 1 is, therefore,
(19)
As follows from the properties of the production functions both sectors are active in t + 1 as
long as
t
< 1. Hence, since individuals are perfectly mobile between the two sectors they
can either supply one unit of labour to the agriculture sector and receive the wage or
GALOR et al. Page 9
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
supply h
t+1
efficiency units of labour to the manufacturing sector and receive the wage
income .
23
Hence,
(20)
and the fraction of employment in the manufacturing sector,
t+1
, equalizes the marginal
product of workers in the two sectors, and thus maximizes output per capita in the economy.
Lemma 1The fraction of workers employed by the manufacturing sector in period t + 1,
t+1
is uniquely determined:
where
X
(y
t
,
t
; X) < 0,
y
(y
t
,
t
; X) > 0, and lim
y
(y
t
,
t
; X) = 1.
Moreover,
t+1
maximizes output in period t + 1, y
t+1
:
Proof: Substituting (3), (5), and (16) into (20) it follows that
(21)
Hence, since (
t+1
, y
t
,
t
; X)/
t+1
> 0, it follows from the Implicit Function Theorem
that there exists a single-valued function
t+1
= (y
t
,
t
; X), where the properties of the
function are obtained noting the properties of the function h(
t
y
t
) and F
L
(X, 1
t+1
).
Moreover, since
t+1
equalizes the marginal return to labour in the two sectors, and since the
marginal products of all factors of production are decreasing in both sectors,
t+1
= argmax
y
t+1
. ||
Corollary 1Given land size, X, prices in period t + 1 are uniquely determined by y
t
and
t
. That is,
Proof: As established in Lemma 1,
t+1
= (y
t
,
t
; X), and the corollary follows noting (3),
(5), (16), and (19). ||
3.4. Efficient expenditure on public education
This section demonstrates that the level of expenditure on public schooling (and hence the
level of taxation) that maximizes aggregate output is optimal from the viewpoint of all
individuals except for landowners who own a large fraction of the land in the economy.
23
Even if mobility between the sectors is not fully unrestricted, the qualitative results would not be altered.
GALOR et al. Page 10
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
Lemma 2Let be the tax rate in period t that maximizes aggregate output in period t +
1,
a. equates the marginal return to physical capital and human capital:
b. is unique, and *(y
t
)y
t
, is strictly increasing in y
t
.
c.
.
d.
.
e.
.
f.
.
g.
.
Proof
a. As follows from (15), (19), and Lemma 1, aggregate output in period t + 1, y
t+1
is
(22)
Hence, since, as established in Lemma 1,
t+1
= (y
t
,
t
; X) = argmax y
t+1
, it
follows from the envelop theorem that
(23)
Furthermore, since then , and thus as follows
from (15),
(24)
Noting 16, satisfies
(25)
and the proof follows, noting that , and .
b. As follows from (24),
(26)
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
Hence, since for all and lim
e0
+ h(e
t
) = , it follows that
for all y
t
> 0. The uniqueness of follows from the properties of
the function . Furthermore, *(y
t
)y
t
is increasing in y
t
. Suppose not.
Suppose that *(y
t
)y
t
is decreasing in y
t
. It follows that * is strictly decreasing in
y
t
, and therefore the L.H.S. of (26) is strictly increasing in y
t
, whereas the R.H.S. is
decreasing, a contradiction.
c.
As derived in part (a), since , it follows from the envelope theorem
that
(27)
d.
Follows from part (c) noting that, as follows from (17), .
e. As follows from part (c),
(28)
and therefore for any
t+1
,
(29)
Moreover, since
(30)
it is strictly increasing in [(1
t
) y
t
]
[h(
t
y
t
)]
1
, and therefore
.
f. As follows from (3) and (20),
(31)
and therefore since w
t+1
is monotonically increasing in
t+1
it follows from part (e)
that .
g.
Follows from part (f) noting that along the factor price frontier
t
decreases in
and therefore in w
t
. ||
As established in Lemma 2 the value of is independent of the size of land, X. The size of
land has two opposing effects on that cancel one another due to the CobbDouglas
production function in the manufacturing sector. Since a larger land size implies that
employment in the manufacturing sector is lower, the fraction of the labour force whose
productivity is improved due to taxation that is designed to finance universal public
education is lower. In contrast, the return to each unit of human capital employed in the
manufacturing sector is higher, while the return to physical capital is lower, since human
capital in the manufacturing sector is scarce.
Furthermore, since the tax rate is linear and the elasticity of substitution between human and
physical capital in the manufacturing sector is unitary, as established in Lemma 2, the tax
rate that maximizes aggregate output in period t + 1 also maximizes the wage per worker,
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
w
t+1
, and the net return to capital, . Hence, there is no conflict of interest among
individuals who do not own land regarding the optimal education policy.
24
Moreover, given
the factor price frontier, since maximizes the wage per worker, w
t+1
, it minimizes the rent
on land,
t+1
.
As follows from Lemma 2, the desirable tax policy from the viewpoint of individual i
depends on the income that the individual derives from land holding, x
i

t+1
, relative to the
income that the individual generates from capital holding and wages, . In
particular, as established in the following proposition, individuals whose land income is
sufficiently small relative to their capital and wage income would support the efficient tax
policy.
Proposition 1Given ( , y
t
, X), there exists a sufficiently low level of land holding by
individual i, , such that the desirable level of taxation from the viewpoint of individual i is
the level of taxation that maximizes output per capita, is inversely related to the level
of .
Proof: Since the indirect utility function is a strictly increasing function of the individuals
second-period wealth, the desirable level of taxation from the viewpoint of individual i
maximizes . As established
in Lemma 2, is maximized at an interior level , and x
i
(y
t
,
t
; X) is minimized by the same interior level . Hence, for all x
i
, is the extremum of
, and thus . In particular for x
i
= 0, is a global
maximum of . Thus, it follows from continuity that there exists such
that for all , the extremum, , remains a global maximum of .
Since , it follows from the continuity of in x
i
that
there exists a sufficiently low level of x
i
, , such that for all (i.e.
there exists a sufficiently low x
i
, such that maximizes globally), where is inversely
related to the levels of .
3.5. The class structure
Suppose that in period 0 the economy consists of three homogeneous groups of individuals
in the first period of their liveslandowners, capitalists, and workers. They are identical in
their preferences and differ only in their initial wealth and landownership. Landowners are a
fraction (0, 1) of all individuals in society who equally share the entire stock of land in
the economy, X. Since landowners are homogeneous in period 0 and since land is
bequeathed from parent to child and each individual has a single child and a single parent, it
follows that the distribution of landownership in society is constant over time, where each
landlord owns X/ units of land. Capitalists are a fraction (0, 1) of all individuals in
society who equally share the entire initial stock of physical capital.
25
Finally, workers are a
24
The absence of disagreement between the capitalists and workers about the optimal tax policy would hold as long as the production
function is CobbDouglas. However, even if the elasticity of substitution was different from 1, in contrast to land owners, both groups
would support public education although they would differ in their desirable tax rates. If the elasticity is larger than unity but finite,
then the tax rate that maximizes the wage per worker would have been larger than the optimal tax rate and the tax rate that maximizes
the return to capital would have been lower, yet strictly positive. If the elasticity of substitution is smaller than unity, the opposite
holds.
25
Heterogeneity in capital holdings across capitalists will not affect the analysis since as established in the discussion that follows
Lemma 2, there is no conflict of interest among the landless. Furthermore, if each landowner, as well, owns an equal stock of capital
in the first period, the qualitative analysis will not be affected.
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
fraction 1 (0, 1) of all individuals in society. They are landless and they do not own
physical capital. Since individuals are initially homogeneous within a group, the uniqueness
of the solution to their optimization problem assures that their offspring are homogeneous as
well. Hence, in every period a fraction of all adults are homogeneous descendants of the
capitalists, a fraction 1 are homogeneous descendants of workers, and a fraction
are landowners. As the economy develops, members of all segments of society accumulate
physical capital.
3.6. Political mechanism
In light of our interest in the effect of economic rather than political transitions on education
reforms and economic growth, the political structure of the economy is designed as a
stationary structure that is unaffected by economic development. In particular, we
deliberately impose a crude political mechanism under which education reforms require the
consent of the class of Landowners. Although economic development does not affect this
political structure, it changes the economic incentives confronted by landowners and thereby
affects their attitude towards educations reforms.
Clearly, even in democracies, the median voting model is not perfectly applicable. Strong
interest groups, such as landowners, exert a larger influence on public policy relative to their
representation in the population. For the sake of simplicity we adopt an extreme modelling
approach that provides landowners as a group with a veto power against education reforms.
The adoption of some alternative approaches, such as a lobbing model, or probabilistic
voting model (Lindbeck and Weibull, 1987), would not change the qualitative results.
Moreover, in order to focus on the conflict between Landowners and the remaining
segments of the economy, we abstract from a potential conflict of interest among
landowners, assuming land is equally distributed across landowners, and coordination
among landowners is therefore not essential.
26
Suppose, in particular, that changes in the existing educational policy require the consent of
all segments of society.
27
In the absence of consensus the existing educational policy
remains intact. Suppose further that consistently with the historical experience, societies
initially do not finance education (i.e.
0
= 0). It follows that unless all segments of society
would find it beneficial to alter the existing educational policy, the tax rate will remain 0.
Once all segments of society find it beneficial to implement educational policy that
maximizes aggregate output, this policy would remain in effect unless all segments of
society would support an alternative policy. Since the landless (i.e. workers and capitalists)
are unified in their support for an efficient level of taxation in every time period, the consent
of the landowners is the pivotal force in the implementation of the output maximizing
education level.
28
26
The introduction of inequality in land-holdings across landowners would not affect the qualitative results. It would have an
ambiguous effect on the timing of education reforms. Large landowners would be expected to suffer a larger loss in rental rents due to
education reforms and would be engaged in more intense lobbying activity to block these reforms, but their force will be diminished
due to their smaller representation within the group of landowners.
27
For simplicity, it is assumed that the decision on the desirable tax rate is taken by the young generation. A more natural assumption
would be to permit the parental generation to choose the desirable level of taxation and thus the resources that would be devoted to the
education of their children. A departure from warm glow utility would achieve this goal at the cost of significant complications. In
particular, if individuals utility is defined over their offsprings income, parents would choose the desirable tax rate from the
viewpoint of the child. This departure would maintain the crucial feature of a monotonic relationship between bequest and income, but
since the total size of transfer will not necessarily be a constant fraction of wealth it would complicate the analysis unnecessarily.
Similarly, the choice of an income tax rather than bequest tax would complicate the analysis. As long as the parental generation
chooses the tax rate on their income, individuals would optimally allocate their income between their own consumption, transfer to
their offspring, and finance of public education. Hence, as long as individuals take the tax structure into account when deciding how
much to bequest, it would not affect the result qualitatively.
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
3.7. Landowners desirable schooling policy
The income of each landowner in the second period of life, , as follows from (8) and
Corollary 1, is
(32)
and , as follows from (10) is therefore
(33)
As summarized in the following Lemma, the economy advances and the share of land in
aggregate output gradually declines, the stake of landowners in other sectors gradually
increases, due to their labour and capital holdings, and their objection to education reforms
therefore declines over time.
29
Proposition 2In the absence of taxation in the initial period, that is,
0
= 0, given the
political mechanism,
a. There exists a critical level of the aggregate capital inheritance of all landowner,
, where , above which their income under the efficient tax policy is
higher than under
t
= 0, and the economy switches to the tax rate that
maximizes income per capita.
b.
The critical level of capital holdings, , above which the efficient tax policy is
chosen,
i. increases with the degree of land inequality in the economy, that is,
ii. is 0 for a sufficiently low level of land inequality (i.e. for a sufficiently
large ). In particular,
28
Landowners, as well as other owners of factors of production, influence the level of public schooling but are limited in their power
to levy taxes for their own benefit. Otherwise, following the Coasian Theorem, the landed elite would prefer an optimal level of
education, taxing the resulting increase in aggregate income.
29
The proposed theory relies on the diminishing importance of land rents for the income of the economy over time, in accordance
with the long-run trend in developed countries. For the U.K., Lindert (1986) documents that the share of land rent in national income
in 1867 was 5%, falling to less than 05% in 19721973. A similar pattern is found for the U.S., where in 1900 the share of national
income going to rent was 91%, by 1930 was 66%, and by 2005 was 07%. (The 1900 figure is from the U.S. Historical Statistics,
series F186191. The 1930 and 2005 figures are from the Bureau of Economic Analysis.) If land is used only in the agricultural
sector, the decline in its rental rate in the process of development, to a level below a positive threshold, assures that landowners would
ultimately support education reforms. If land is also used in the manufacturing sector, the results will not be affected qualitatively, as
long as the share of land that is employed in the industrial sector is initially small. The rise in the rental rate on industrial land in the
process of urbanization and its impact on the rise on the rental rate of land in the economy as a whole, would just accelerate the
transition, since it will increase landowners benefits from the process of industrialization.
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
iii. is 0 for a sufficiently large level of income per capita. In particular,
iv. Let t
be the first period in which the efficient tax policy, , is

implemented. The efficient tax policy will remain in place thereafter, that
is,
a. Noting that landlords are identical and their number is unchanged in the process of
development, the tax policy that maximizes income of all landowners also
maximizes the income of each landowner. As follows from (32), is
the level of that equates the income of landowners in the case were
t
=0
and exists since as established in Lemma 2 , and
thus .
b.
i. Follows directly from the derivation of b
(y
t
; X, ), with respect to ,
noting that for a given y
t
, has no effect on prices and that for y
t
> 0,
, and therefore .
ii. Since the agriculture production function (2) is CRS, it follows that the
aggregate return to land is
(34)
Hence, landlords income, , is
(35)
Since
t+1
= argmax
t
= argmax F(X, 1
t+1
) w
t+1
(1
t+1
), it
follows from the envelope theorem that
(36)
Thus if > 1
t+1
> 0, it follows from Lemma 2 that for any ,
(37)
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
and therefore for a sufficiently large the threshold is 0, that is, lim
1
B
L
(y
t
; X, ) 0.
iii. As follows from Lemma 1, as y
t
,
t+1
1 and therefore it follows
from (36) that for any , (37) holds and hence lim
y
t

B
L
(y
t
; X, )
0.
c. As established in Proposition 1, the desirable tax policy from the viewpoint of
landless (i.e. workers and capitalists) is . Hence, given that the political
mechanism requires a consensus for changes in the tax policy, once the chosen tax
rate is it will remain so thereafter.
30
||
Remark 1There exists a range of agricultural production functions such that the desirable
level of taxation from the viewpoint of landowners, , are
t
= 0 or , in the range
.
31
It should be noted that given the political mechanism, and the absence of
taxation in period 0, even if the desirable level of taxation from the viewpoint of a
landowner, , is any level in the interval (0, ), the tax rate that prevails in the economy in
every period t is either 0 or . Under a different political structure the transition from a zero
tax rate to could be a gradual process. The process of development will induce
landowners to compromise (or support) increasingly higher levels of taxation and the
qualitative results regarding the adverse effect of land inequality on the implementation of
education reforms would remain intact.
4. THE PROCESS OF DEVELOPMENT
This section analyses the evolution of an economy from an agricultural to an industrial-
based economy. It demonstrates that the gradual decline in the importance of the agricultural
sector along with an increase in the capital holdings in landlords portfolio may alter the
attitude of landlords towards educational reforms. In societies in which land is scarce or its
ownership is distributed rather equally, the process of development allows the
implementation of an optimal education policy, and the economy experiences a significant
investment in human capital and a rapid process of development. In contrast, in societies
where land is abundant and its distribution is unequal, an inefficient education policy will
persist and the economy will experience a lower growth path as well as a lower level of
output in the long-run. Thus, land reforms that sufficiently reduce inequality in
landownership permit an earlier implementation of an efficient education policy.
Proposition 3The conditional evolution of output per capita, as depicted in Figure 1, is
given by
30
It should be noted that, in fact, landowners optimal tax rate will remain thereafter, since education reforms would further
increase the stake of landowners in the non-agricultural part of the economy.
31
In particular, the preferred tax rate from the viewpoint of landowners will be
t
= 0 or when the elasticity of substitution
between labour and land is 0 or 1. (i) If the production function is CobbDouglas , as established in Appendix
A, landowners would prefer either
t
= 0 or over any . (ii) If land and labour are perfect complements, as
established in Proposition 5, as long as the wage rate is below the threshold level above which the demand for workers in agriculture is
0, landowners prefer the lowest level of industrial output, , and hence
t
= 0. As the economy develops and the wage rate crosses
this threshold, their preferred tax rate is since the return to land is 0 anyway.
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
where,
Proof: As follows from (1), (15), and (19),
. Thus, noting that, h(0) = 1 the
evolution of y
t+1
as stated in the proposition is obtained. Since and , it
follows that *(y
t
) >
0
(y
t
) for y
t
> 0. As follows from Lemma 1 and Proposition 2, the
properties of the functions *(y
t
) and
0
(y
t
) follows, noting that
t+1
= argmax y
t+1
,
and applying the envelop theorem. ||
Note that the evolution of output per capita, for a given schooling policy, is independent of
the distribution of land and income.
Corollary 2Given the size of land, X, there exists a unique
0
and a unique * such that
where * >
0
.
Proof: Follows from the properties of *(y
t
) and
0
(y
t
), as established in Proposition 3. ||
The evolution of income per capita, as depicted in Figure 1, and as follows from
Propositions 2 and 3, is
Hence, the economy evolves on the lower trajectory dictated by
0
(y
t
) till time t
(e.g. where
the level of income is y
t
) and then moves to a higher trajectory that is governed by
*(y
t
).
Proposition 4For a given set of initial conditions, (i.e. y
0
, k
0
, X, h
0
= 1,
and therefore
0
= 0), a less equal land distribution (i.e. a low level of
) will generate a delay in the implementation of an efficient education policy and will
therefore result in an inferior growth path. That is, a less equal distribution of landownership
implies that the timing of the implementation of the efficient tax policy, t
, is delayed.
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
Proof: As follows from (33), noting that , the evolution of aggregate
capital holdings of landowners, for
t
= 0, and for t > 0 is
(38)
As established in Proposition 3, as long as
t
= 0, the evolution of income per capita, y
t
, is
independent of . Hence it follows from Corollary 1 that factor prices are independent of
and therefore, as follows from (38), is increasing in . Hence, noting that as established
in Proposition 2 B
L
(y
t
; X, ) is decreasing with , the lower is the larger is t
(i.e. the later

is the time period in which ).
Proposition 5 (Persistence of Inefficient Education Policy)If the productivity in
the manufacturing sector is limited, and the degree of complementarity between land and
labour is sufficiently high, then there exists a sufficiently high level of land inequality (i.e. a
sufficiently low ), such that inefficient education policy will persist indefinitely (i.e. t

).
Proof: Suppose that the production function in the agriculture sector is
, where, X < 1 (i.e. X is smaller than the size of the working
population) to assure that some workers are employed in the industrial sector. Hence, for w
t
< 1, X = L
t
= 1
t
. As follows from (18) and (20), . Therefore, for w
t
< 1,
Suppose, for the sake of simplicity, that X = . Then, for
Hence, if for , the income of all landowners, noting (17), is
where is the share of landowners in the total capital stock. Since , it follows that for a
sufficiently low landowners income is decreasing with , as long as . Hence, since
0
< *(
0
), then if *(
0
) < 1 landowners prefer
t
= 0, rather than when y
t
=
0
. ||
Corollary 3 (Land Reforms and Education Policy)A land reform that reduces
suffi-ciently the concentration of landownership in the economy (i.e. a sufficient increase in
) would expedite the implementation of efficient education policy.
Proof: Follows from Proposition 4. ||
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
Hence, consistent with historical anecdotes presented in the next section, land reforms
would be expected to follow by education reforms.
Under the conditions specified in Proposition 5 there exists a steady-state equilibrium in
which an inefficient education policy exists. In particular, as depicted in Figure 2, country A
reaches a steady-state equilibrium at a level of income per capita [
0
]
A
, prior to the
implementation of education reforms that would have occurred if the level of income per
capita in the economy would have reached
A
.
Thus, among countries where land inequality is higher (i.e. is smaller) a poverty trap, in
which inefficient education policy persists may emerge. In particular, a country could reach
the low income steady state
0
before reaching the point in which is sufficiently large to
bring about a policy shift. In contrast, for sufficiently equal economies, t
is necessarily
finite. In particular if landownership is equally distributed across members of society (i.e. if
= 1), then as established in Proposition 2, the efficient tax policy is implemented in period
0.
Hence, the distribution of land within and across countries affected the nature of the
transition from an agrarian to an industrial economy, generating diverging growth patterns
across countries. Furthermore, land abundance that was beneficial in early stages of
development, brought about a hurdle for human-capital accumulation and economic growth
among countries that were marked by an unequal distribution of landownership. As depicted
in Figure 2, some land-abundant countries, which were associated with the club of the rich
economies in the pre-industrial revolution era and were characterized by an unequal
distribution of land, were overtaken in the process of industrialization by land-scarce
countries. The qualitative change in the role of land in the process of industrialization has
brought about changes in the ranking of countries in the world income distribution.
32
In the process of development, as long as the economy implements an efficient education
policy, inequality subsides over time. In particular, inequality between workers and
capitalists asymptotically disappear, whereas inequality between landowners and the
landless subsides, due to the decline in the return to land. If the economy remains in a
poverty trap, however, inequality between landowners and landless will not converge, while
inequality between workers and capitalists will asymptotically disappear.
33
Land inequality and wealth inequality may have a very different effect on education
reforms. While inequality in landownership delays education reforms, inequality in the
distribution of capital among the landless has no effect on the timing of education reforms,
whereas a larger concentration of capital held by the landowners would expedite the
implementation of education reforms.
32
If the utility of individuals is defined over the discounted stream of utilities of their offspring, the qualitative results will not be
affected. An earlier implementation of education reforms would raise the income of future members of a landowners dynasty on the
account of the contemporary income of the landowner. The optimal timing of the implementation of education reforms from the
viewpoint of each landowner would depend, therefore, on the discount factor applied for future members of the dynasty. It would
occur earlier than in the case in which individuals do not generate utility from the utility of their offspring, but would be still affected
adversely by the degree of land inequality, since it determines the relative stake of landowners in other segments of the economy. In
particular, if * < 1, in the context of Proposition 5, there exists a sufficiently high level of land inequality such that inefficient
education policy will persist indefinitely (i.e. landowners would not find it beneficial to implement education reforms in any time
period). In this case, regardless of the discount factor applied to offspring the timing of education reforms will not be affected at all
(i.e. t
).
33
The distinction between workers and capitalists fades in the limit due to the simplifying structure of homothetic preferences. If
preferences are non-homothetic, as in Galor and Moav (2006), inequality in the distribution of landownership, that would delay the
implementation of efficient schooling would slow down this convergence process.
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
5. HISTORICAL EVIDENCE
Historical evidence suggests that indeed the distribution of landownership has been a
significant force in the emergence of sustained differences in human-capital formation and
growth patterns across countries.
5.1. Landownership and the level of education
Anecdotal evidence suggests that the degree of concentration of landownership across
countries and regions is inversely related to education expenditure and attainment. North and
South America provide the most distinctive set of suggestive evidence about the relationship
between the distribution of landownership, education reforms, and the process of
development. The original colonies in North and South America had a vast amount of land
per person and levels of income per capita that were comparable to the Western European
ones. North and Latin America, however, differed in the distribution of land and resources.
While the U.S. and Canada have been characterized by a relatively egalitarian distribution of
landownership, in the rest of the new world land and resources have been persistently
concentrated in the hand of the elite (Deninger and Squire, 1998).
Consistent with the proposed theory, persistent differences in the distribution of
landownership between North and Latin America were associated with significant
divergence in education and income levels across these regions (Maddison, 2001). Although
all of the economies in the western hemisphere were developed enough in the early 19th
century to justify investment in primary schools, only the U.S. and Canada were engaged in
the education of the general population (Coatsworth, 1993; Engerman and Sokoloff, 2000).
34
Variations in the degree of inequality in the distribution of landownership among Latin
American countries were reflected in variation in investment in human capital as well. In
particular, Argentina, Chile, and Uruguay, in which inequality in the distribution of
landownership was less pronounced, invested significantly more in education (Engerman
and Sokoloff, 2000). Similarly, Nugent and Robinson (2002) show that in Costa Rica and
Colombia, where coffee is typically grown in small farms (reflecting lower inequality in the
distribution of land) income and human capital are significantly higher than that of
Guatemala and El Salvador, where coffee plantations are rather large.
35
Moreover, one of
the principles championed by the Progressives during the Mexican Revolution of 1910 was
compulsory, free public education. However, the achievement of this goal varied greatly by
state. In the north, with a relatively more equitable land distribution, enrolment in public
schools increased rapidly as industrialization advanced following the revolution. This is in
contrast to the southern states, which were dominated by the haciendas who employed
essentially slave labour. In these states there was virtually no increase in school enrolment
following the revolution (Vaughan, 1982). Similarly, rural education in Brazil lagged due to
the immense political power of the local landlords. Hence, in 1950, 30 years after the
Brazilian government had instituted an education reform, nearly 75% of the nation was still
illiterate (Bonilla, 1965).
34
One may view the civil war in the U.S. as a struggle between the industrialists in the north who were striving for a large supply of
(educated) workers and the landowners in the south who wanted to sustain the existing system and to assure the existence of a large
supply of cheap (uneducated) labour.
35
In contrast to the proposed theory, Nugent and Robinson (2002) suggest that a hold-up problem generated by the monopsony
power in large plantations prevents commitment to reward investment in human capital, whereas smallholders can capture the reward
to human capital and have therefore the incentive to invest. This mechanism does not generate the economic forces that permit the
economy to escape this institutional trap.
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
5.2. Land reforms and education reforms
Evidence from Japan, Korea, Russia, and Taiwan indicates that land reforms were followed
by, or occurred simultaneously with, significant education reforms. There are two
interpretations for those historical episodes. First, as proposed explicitly by the theory, land
reforms may diminish the economic incentives of landowners to block education reforms.
Second, a non-favourable shift in the balance of power from the viewpoint of the landed
aristocracy brought about the implementation of both land and education reforms,
consistently with the basic premise that landowners opposed education spending whereas
others (e.g. the industrial elite) favoured it.
5.2.1. Japan and the Meiji restorationTowards the end of Tokugawa regime (1600
1867), although the level of education in Japan was impressive for its time, the provision of
education was sporadic and had no central control or funding, reflecting partly the resistance
of the land-holding military class for education reforms (Gubbins, 1973). The opportunity to
modernize the education system arrived with the overthrow of the traditional feudal
structure shortly after the Meiji Restoration of 1868. In 1871, an Imperial Decree initiated
the abolishment of the feudal system. In a sequence of legislation in the period 18711883,
decisions on land utilization and choice of crops were transferred to farmers from their
landlords, prohibitions on the sale and mortgage of farmland were removed, a title of
ownership was granted to the legal owners of the land, and communal pasture and forest
land was transferred from the ownership of wealthy landlords to the ownership of the central
government. This legislation resulted in the distribution of land holdings among small
family farms, which persisted until the rise of a new landlord system during the 1930s
(Hayami, 1975, ch. 3).
Education reform and land reform evolved simultaneously. In 1872 the Educational Code
established compulsory and locally funded education for all children between ages 6 and 14
(Gubbins, 1973, ch. 30). In addition, a secondary school and university system was funded
by the central government. The Education Code of 1872 was refined in 1879 and 1886,
setting the foundations for the structure of Japanese education until World War II. The
progress in education attainment following the land reforms of the Meiji government was
substantial. While in 1873 only 28% of school-age children attended schools, this ratio
increased to 51% by 1883 and to 94% by 1903 (Passin, 1965).
5.2.2. Russia before the revolutionEducation in Tsarist Russia lagged well behind
comparable European countries at the close of the 19th century. Provincial councils
dominated by wealthier landowners were responsible for their local school systems and were
reluctant to favour education for the peasants (Johnson, 1969). Literacy rates in the rural
areas were a mere 21% in 1896, and the urban literacy rate was only 56%. As the Tsars grip
on power weakened during the early 1900s the political power of the wealthy landowners
gradually declined leading to a sequence of agrarian reforms that were initiated by the
premier Stolypin in 1906. Restrictions on mobility of peasants were abolished, fragmented
land-holdings were consolidated, and the formation of individually owned farms was
encouraged and supported through the provision of government credit. Stolypins reforms
accelerated the redistribution of land to individual farmers and land-holdings of the landed
aristocracy declined from about 3545% in 1860 to 17% in 1917 (Florinsky, 1961).
Following the agrarian reforms and the declining influence of the landed aristocracy, the
provision of compulsory elementary education had been proposed. The initial effort of 1906
languished, but the newly created representative Duma continued to pressure the
government to provide free compulsory education. In the period 19081912, the Duma
approved a sequence of significant increases in expenditures for education (Johnson, 1969).
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
The share of the Provincial councils budget that was allocated to education increased from
204% in 1905 to 311% in 1914 (Johnson, 1969), the share of the central governments
budget that was devoted to the Ministry of Public Education increased three-fold from 14%
in 1906 to 49% in 1915, and the share of the entire population that was actively attending
schools increased threefold from 17% in 1897 to 57% in 1915 (Dennis, 1961).
5.2.3. South Korea and TaiwanThe process of development in Korea was marked by
a major land reform followed by a massive increase in governmental expenditure on
education. During the Japanese occupation in the period 19051945, land distribution in
Korea became increasingly skewed, and in 1945 nearly 70% of Korean farming households
were simply tenants (Eckert, 1990). In 19481950, the Republic of Korea instituted the
Agricultural Land Reform Amendment Act that drastically affected land-holdings.
36
The
principle of land reform was enshrined in the constitution of 1948 and the actual
implementation of the Agricultural Land Reform Amendment Act began on 10 March
1950.
37
This act prohibited tenancy and land renting, put a maximum on the amount of land
any individual could own, and dictated that an individual could only own land if they
actually cultivated it. Owner cultivated farm households increased sixfold from 349,000 in
1949 to 1,812,000 in 1950, and tenant farm households declined from 1,133,000 in 1949 to
essentially 0 in 1950 (Yoong-Deok and Kim, 2000).
Land reforms were accompanied by soaring expenditure on education. In 1949, a new
Education Law was passed within South Korea that focused specifically on transforming the
population into a technically competent work-force capable of industrial work. This led to
dramatic increases in the number of schools and students at all levels of education. Between
1945 and 1960 the number of elementary schools increased by 60% and the number of
elementary students went up by a staggering 165%. In secondary education the growth is
even more dramatic, with both the number of schools and the number of students growing
by a factor of ten in the same time period. The number of higher education institutions
quadrupled and the number of higher education students increased from only 7000 in 1945
to over 100,000 in 1960. In 1948, Korea allocated 8% of government expenditure to
education. Following a slight decline due to the Korean war, educational expenditure has
increased to 92% in 1957 and 149% in 1960, remaining at about 15% thereafter (Sah-
Myung, 1983).
Taiwan experienced a similar path over the same period once the Japanese colonization
ended. The government of Taiwan implemented a land reform in the time period 1949
1953. It enforced rent reductions, sold public land to individual farmers who had previously
been tenants, and permitted the purchase of rented land. In 1948, prior to the land reform,
57% of farm families were full or part owners, 43% were tenants or hired hands. By 1959
the share of full or part owners had increased to 81%, and the share of tenants or hired hands
dropped to 19% (Cheng, 1961).
A massive educational reform accompanied these land reforms. The number of schools in
Taiwan grew by 5% per year between 1950 and 1970, while the number of students grew by
6% a year. The pattern of growth mirrors that of South Korea, with especially impressive
growth of 11% per year in the number of secondary students and 16% per year in the
36
A major force behind this land reform was the aim of the U.S. provisional government after World War II to remove the influence
of the large landowners (who were either Japanese or collaborators with the Japanese).
37
Formally the education reform took place prior to the land reforms, but the provision for land reform was enshrined in the
constitution prior to the educational reform. The imminent land reform could have reduced the incentives for the landed aristocracy to
oppose this education reform.
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
number of higher education students. Funding for education grew from 178% of GNP in
1951 to 412% in 1970 (Lin, 1983).
In 1950 South Korea and Taiwan were primarily agricultural economies with a GDP per
capita (measured in 1990 international dollars) of $770 and $936, respectively. South Korea
and Taiwan lagged in GDP per capita well behind many countries within Latin America,
such as Colombia ($2153) and Mexico ($2365), sharing with these countries a legacy of vast
inequality in the distribution of agricultural land. In contrast to the Latin American
countries, the implementation of land reforms in South Korea and Taiwan and its apparent
effect on education reforms affected their growth trajectory significantly, leading them to
one of the most successful economic growth stories of the post-war period. From a level of
income per capita in 1950 that placed them not only far behind the nations of Latin America
but also behind Congo, Liberia, and Mozambique, these two countries have each grown at
an average rate of nearly 6% per year between 1950 and 1998, leaving behind the countries
of sub-Saharan Africa and overtaking the Latin American countries in this period. In 1998
South Korea and Taiwan had GDP per capita levels 150% higher than Colombia and 100%
higher than Mexico (Maddison, 2001).
6. EVIDENCE FROM THE U.S. HIGH-SCHOOL MOVEMENT
The central hypothesis of this research, that land inequality adversely affected the timing of
education reforms, is examined empirically using variations in public spending on education
across states and over time in the U.S. during the high-school movement. Historical
evidence from the U.S. on education expenditures and landownership in the period 1880
1940 suggests that land inequality had a significant adverse effect on educational
expenditures during this period.
38
During the first half of the 20th century the education system in the U.S. underwent a major
transformation from insignificant to nearly universal secondary education. As established by
Goldin (1998), in 1910 high-school graduation rates were between 9 and 15% in the
Northeast and the Pacific regions and only about 4% in the South. By 1950 graduation rates
were nearly 60% in the Northeast and the Pacific regions and about 40% in the South.
Furthermore, Goldin and Katz (1997) document significant variations in the timing of these
changes and their extensiveness across regions.
The high-school movement and its qualitative effect on the structure of education in the U.S.
reflected an educational shift towards non-agricultural learning that is at the heart of the
proposed hypothesis. The high-school movement was undertaken with the intention of
building a skilled work-force that could better serve the manufacturing sector. Over this
period, firms increasingly demanded skilled workers who could be effective managers, sales
personnel, and clerical workers, and courses in accounting, typing, shorthand, and algebra
were highly valued in the white-collar occupations. In addition, in the 1910s, some of the
high-technology industries of the period started to demand blue-collar craft workers who
were trained in mathematics, chemistry, and electricity (Goldin, 1999).
The proposed theory suggests that inequality in the distribution of landownership was
significant in determining the pace of education reforms across the U.S. We exploit
differences in education expenditures across states over the period 19001940 to identify the
role of the land inequality on education expenditures, controlling for the level of income per
capita, the percentage of the black population, and the urbanization rate within each state.
39
38
For other studies of the relationship between land and economic performance in the U.S. over this time period see Gerber (1991)
and Caselli and Coleman (2001).
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
6.1. Data
The level of expenditure per child within each state in the time period 19001940 is
computed, utilizing data on the number of children in the state in each of the relevant years
from the relevant U.S. Census.
40
These expenditures are converted to 1920 dollars to match
the income per capita estimates used.
41
Education expenditure varied greatly over this period. For example, in 1900 the state of
Alabama was spending $316 (in 1920 dollars) per child on education. In contrast,
Massachusetts had expenditures of $4457 per child, a 14-fold difference. By 1920, Alabama
had expenditures of $1178 per child, while spending per child in Massachusetts had
increased to $4509, only four times greater than Alabamas spending. In 1940, the gap had
narrowed to less than a factor of three, $3561 for Alabama and $10287 for Massachusetts.
The degree of inequality in the distribution of landownership is captured in a consistent
fashion with the structure of the model by the share of land held by large landowners. In
particular, based on U.S. Census data, we trace the evolution of the share of land holdings
by the minimal number of farms that constitute 20% of agricultural land in 1880 within each
state, as outlined in Appendix B. For subsequent years, 1900 and 1920, the share of land
held by this same number of the largest farms is measured. To illustrate the methodology, in
Wisconsin in 1880 the largest 15,145 farms (11% of total farms) held 20% of the farmland.
In 1900, the largest 15,145 farms held 16% of the land, declining to 12% in 1920. The
qualitative results are not affected if we use alternatively as a benchmark the share of land
holdings by the minimal number of farms that held 5%, 10%, 25%, 50%, or 80% of the land
in 1880.
The evolution of land concentration across regions in the U.S. (as defined in Appendix C)
exhibits the following patterns. For states in the Northeast, the average share rose from 20%
in 1880 to 22% in 1900 and 24% in 1920. Southern states experienced a decline in the
average share of land held by the largest farms from 20% in 1880 to 12% in 1900 and to
only 8% by 1920. This decline in the share of land held by large farmers is mimicked in the
West, where the share drops to 9% in 1900 and to 6% in 1920. Similarly, the Midwest
experienced a decline from 20% in 1880 to 16% in 1900 to 13% in 1920.
Several other controls are included in the specifications. Variation in income per capita
across states that would be expected to affect the variation in education expenditures. The
percentage of the black population is included to ensure that the adverse effect of inequality
in the distribution of landownership on educational expenditure does not reflect the adverse
effect of the discrimination in the South (where land inequality is more pronounced), on
educational expenditure.
42
The final control, the percentage of urban population is added for
several reasons. Given economies of scale, it may be that more urbanized states in fact have
lower expenditures per child due to their higher density. Furthermore, urbanization and
industrialization are highly correlated, and urbanization may partly control for capital
intensity across states and the higher demand for human capital in the urban sector.
43
39
Consistently with the proposed theory and the empirical findings, Wright (1986) suggests that Southern governments, influenced
heavily by land-holders, refused to expand enrolments and spending in education because the North which provided a significant
outside option for educated workers would reap the benefits from it.
40
The precise age ranges used in each census vary, but as these changes are common to all states, this does not introduce any bias
into the results. The available age ranges are 517 years in 1880, 520 years in 1900, 720 years in 1920, and 519 years in 1950
41
The total population of school-age children in each state, rather than the actual number of students, is used because states could
control their total expenditures by limiting the number of actual pupils (e.g. the exclusion of blacks from public education in the South
during much of the period under study).
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
6.2. Empirical specifications and results
The empirical analysis examines the effect of inequality in the distribution of landownership
(i.e. the land share of large landowners) in state i in period t 1, on log expenditure per
child in state i in period t, lne
it
, over four periods of observation: 1880, 1900, 1920, and
1940.
44
In particular, for t = 1900, 1920, 1940, and t 1 = 1880, 1900, 1920, respectively,
(39)
where S
i,t1
is the land share of large farms in state i in period t 1, ln y
i,t1
is the log
income per capita in state i in period t 1, U
i,t1
is the percentage of the urban population in
state i in period t 1, B
i,t1
is the percentage of the black population in state i in period t
1, and
it
is the error term of state i in period t 1. This formulation captures the existence
of a lag between the current economic conditions and their effect on the political structure
and the implementation of educational policy.
There are several concerns in exploiting cross-states variations in the distribution of
landownership, and education expenditures to assess the effect of inequality in the
distribution of landownership on education expenditure.
First, an unobserved factor at the state level, which is correlated with the distribution of
landownership, may have affected education expenditure. In order to overcome this concern
we examine the first difference of equation (39), and estimate the effect of changes in land
concentration on changes in education expenditures. This strategy permits the estimation of
the parameter of interest,
1
, while allowing for a time invariant unobserved heterogeneity
across states in the level of the log expenditure per child.
Second, an unobserved factor at the state level may have affected both the changes in the
distribution of landownership and the changes in education expenditure. Our empirical
strategy allows for linear unobserved heterogeneity across states in the time trend of log
expenditure per child. Thus, the estimation of the effect of changes in land concentration on
changes in education expenditures with state fixed effects, controls for a linear unobserved
heterogeneity across states in the time trend of the log expenditure per child. Namely, we
presume that changes in explanatory variables are not correlated with changes in the error
term, even if the levels of the explanatory variables might be correlated with it. Although we
42
Black students often suffered not only from insufficient funding, but were also excluded from the education system entirely in many
places. Margo (1990) identifies several avenues along which black students suffered in relation to their white peers during the periods
of the study. Blacks also lived predominantly in the South, where land inequality was relatively high as a result of the plantation
system. An additional avenue of influence for the black population (and labour in general) involves mobility. Wright (1970) argued
that some Southern states limited education spending because of the fear that the educated workers would migrate out of their home
states. However, while the amount of internal migration was large in absolute terms, relative to the size of the population it was much
less important. Eldridge and Thomas (1957) calculate an index of interstate redistribution, which measures the percentage of the
population that would have to be moved in any decade in order to match the previous decades distribution by state. In 19001910,
this index is 425%, and then is lower in every decade through 19401950. As this index also reflects changes in population
distribution due to fertility differences between states, it overestimates the effect of internal migration. It thus seems likely that there
was appreciable friction to labour mobility, and that local education expenditures could, to some extent, benefit local populations.
Including net migration rates from Eldridge and Thomas (1957) as part of the empirical specifications that follow does not alter the
results.
43
The theory predicts that the size of the capital stock interacts with inequality in landownership to determine the nature of education
expenditure. While the measures of the aggregate capital stock per person by state is available, the inferences of the theory are about
capital holdings by landowners, that is unavailable. Regardless, inclusion of the (log) aggregate capital stock per person in place of, or
in addition to, the urban percentage does not alter any of the empirical results that follow. Moreover, the use of income per capita
controls for some of the effect of capital per worker as well.
44
An alternate specification would be to examine the effect on the log of total expenditure ln E
it
as opposed to the log of expenditure
per child lne
it
. This would eliminate any concern that expenditures per child were changing due to random fluctuations in the size of
the population. Regressions using ln E
it
as the dependent variable, and including the size of the log child population, as an explanatory
variable, do not alter the results qualitatively.
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
do not control for a non-linear state trend, as will become apparent, the coefficient of interest
is robust to the inclusion of the linear time trend, and thus one should not expect the non-
linear specification to affect the significance of the results.
Third, one may be concerned about potential reverse causality from education expenditure to
inequality in landownership. This concern is addressed in several dimensions. We regress
education expenditure, e
it
, on lagged concentration of landownership, S
i,t1
. The lagging
allows for some control of potential reverse causality running from education expenditures
to the land share of the largest farms (i.e. it is reasonable that S in 1900 will affect e in 1920,
but unlikely that S in 1920 will affect e in 1900). However, we might still capture reverse
causation if there is serial correlation in education expenditures. This concern is handled in
three different ways: (a) Since we estimate the effect of changes in land concentration on
changes in education expenditures, the concern is the presence of serial correlation in the
changes in education expenditure. However, we find that there is no serial correlation in the
difference of education expenditures (while there is serial correlation in the level of
education expenditures). (b) We control for state-specific time trends in education
expenditure. (c) An instrument for concentration of landownership is developed that
provides us with exogenous variation in the concentration of landownership, S
t1
, and
permits us to establish the causal effect of land inequality on education expenditures.
Thus, we allow for a time invariant unobserved heterogeneity across states in the level of the
log expenditure per child,
i
, and a linear unobserved heterogeneity across states in the time
trend of the log expenditure per child,
i
t, as well as variations in the time effect at the
national level,
t
. Namely,
(40)
First differencing (39) and utilizing (40) it follows that
(41)
where lne
it
lne
it+1
lne
it
(i.e. the difference in the log expenditure per child in state i
between 1920 and 1900 and between 1940 and 1920), S
i,t1
S
i,t
S
i,t1
(i.e. the
difference in land share of large farms in state i between 1900 and 1880 and between 1920
and 1900), and
t1
=
t

t1
. The lag operator is similarly defined for the rest of the
explanatory variables. Given the empirical specification (41) and the available data, we have
two possible observations for each state. Due to limitations in the data we have 79 total
observations over 41 states, with three states having data only from 1920 and 1940.
The negative correlation between the changes in the log of education expenditure in state i,
lne
it
, and the lagged changes in land share of large farms in state i, S
i,t1
, is apparent in
Figure 3 and is demonstrated by the fitted values plotted from an OLS regression.
Table 1 depicts the correlation between all variables utilized in the empirical specifications,
in particular, the correlation coefficient between lne
it
and S
i,t1
, as depicted in Figure 3.
The table indicates that lagged changes in the land share of large farms are negatively
related to changes in education expenditures per child in the next period. Moreover, changes
in education expenditures are positively associated with the lagged changes in log income
per capita and negatively with the lagged changes in the percentage of the black population.
To undertake more rigorous empirical testing, we begin by assuming that E(
it
) = 0 and
E(
it
X) = 0, where X (S
i,t1
, ln y
i,t1
, U
i,t1
,B
i,t1
). In other words, we presume that
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
the changes in explanatory variables are not correlated with changes in the error term, even
though the levels of the explanatory variables might be correlated with the error term itself
in (39). In addition, we begin by assuming that the time trend parameter,
i
, is identical
across states. Under these assumptions the specification in (41) can be estimated by OLS,
with S.E.s adjusted for clustering by state, allowing for the differenced error terms for state
i,
it
, to be correlated across different time periods.
Table 2 depicts the results of this estimation in columns (1)(5), establishing the negative
effect of the lagged change in land share of large farms, S
i,t1
on the change in log
education expenditure per capita lne
it
, alone and while controlling for the change in
lagged income per capita, ln y
i,t1
, the lagged change in the percentage of the urban
population, U
i,t1
, and the lagged change in percentage of the black population, B
i,t1
.
As indicated by the results in column (1) the effect of S
i,t1
alone on the change in
education expenditure, lne
it
, is negative and highly significant. One would also expect that
changes in education expenditures would reflect changes in income per capita. Controlling
for the change in lagged income per capita, ln y
i,t1
in column (2), shows that indeed an
increase in lagged income per capita has a highly significant, positive effect on education
expenditure. Nevertheless, the negative effect of S
i,t1
on the change in education
expenditure, lne
it
, remains stable and highly significant. Column (3) includes a control for
the lagged change in percentage of the black population, B
i,t1
, to ensure that the adverse
effect of inequality in the distribution of landownership on educational expenditure does not
reflect the adverse effect of the discrimination in the South (where land inequality is more
pronounced), on educational expenditure. As expected the effect of the change in the
percentage of the black population on the change in educational expenditure is negative and
highly significant. However, the effect of the change in the distribution of landownership
remains negative and highly significant. Finally column (4) adds a control for the lagged
change in the percentage of the urban population, U
i,t1
, capturing a potential adverse
effect of urbanization on education expenditure due to economies of scale in education and
its positive effect stemming from the correlation between industrialization and urbanization.
The effect of the lagged changes in urbanization on changes in education expenditure is
negative but insignificant. The negative effect of S
i,t1
on the change in education
expenditure remains stable, negative, and highly significant.
Hence, as follows from (39) and (41) the coefficient
1
that measures the effect of the
lagged change in land share of large farms, S
i,t1
, on the change in log education
expenditure per capita lne
it
, captures the effect of the lagged land share of large farms,
S
i,t1
on log education expenditure per capita lne
it
. The size of the point estimate for S
i,t1
is
relatively stable over the first five specifications, suggesting that a 10 percentage point
decline in S
i,t1
would have increased expenditure per child at the following period by 21
27%. In particular, consider the difference between the land share of large farms in
California and Vermont in 1920. In California S
1920
= 0096 (which is at the 25th percentile
of the distribution of S across states in the U.S.) and in Vermont S
1920
= 0215 (which is at
the 75th percentile). Using the estimates in column (4) this implies that Vermonts
expenditure per child in 1920 would have been 25% higher if it had a land share of large
farms as small as Californias. That difference would have eliminated more than a third of
the actual gap in expenditure per child that existed between California ($68 per child) and
Vermont ($41 per child) in 1940. Column (5) establishes that the inclusion of a common
time trend for all states does not affect the qualitative results, increasing slightly the absolute
value of the point estimate on S
i,t1
.
Column (6) reports the estimation of equation (41) using state fixed effects, allowing for the
time trend in education expenditures,
i
, to vary by state, where all control variables are
included. In comparison to columns (4) and (5) the absolute value of the point estimate of
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
the effect of the change in the lagged land share of large farms has increased, but it is
significant only at the 10% level, reflecting the reduction in the degrees of freedom. A rise
in S.E. and a decline in significance is also observed in all other explanatory variables. The
results in column (6) may seem to indicate that there is some state-specific time trend and
that previously the change in the land share of large farms was a proxy for this state-specific
effect. However, the Hausman test, report in column (6) of Table 2, comparing the fixed
effects estimation to a random effects estimation, indicates that we cannot reject the null
hypothesis that the two specifications differ only randomly.
45
Since the model abstracts from inequality among landowners, the use of the explanatory
variable, S (i.e. the land share of large farms), is appropriate. Nevertheless, it should be
noted that the use of the Gini coefficient for farm size (i.e. land inequality among
landowners), calculated using the same raw data used in creating the variable S, would not
affect the qualitative results. In particular, if S is replaced by the Gini coefficient for
inequality in landownership, land inequality still has an adverse effect on education
expenditure. Furthermore, if both measures are used jointly, then the coefficient of land
share of large farms remains negative (215) and significant at the 1% level, and the effect
of the Gini coefficient also remains negative although insignificant.
6.3. Instrumental variables estimation
This section introduces an instrumental variables analysis to further enhance confidence
about the identification of the effect of the concentration of landownership on education
expenditure. In order to identify the effect of the concentration of landownership on
education, we require a source of exogenous variation in the concentration of landownership
that does not influence education spending directly. In light of the historical evidence
provided by Engerman and Sokoloff (2000) regarding the positive effects of agricultural
crops associated with economies of scale (e.g. cotton and sugar cane) on land inequality
across the Western Hemisphere, one should expect that cross-state differences in climatic
characteristics, and thus in the suitability for such crops, would generate variation in the
concentration in landownership across states. Moreover, nationwide changes in the relative
prices of agricultural crops that are associated with economies of scale would generate
changes in the concentration of landownership over time. Thus, the interaction between
nationwide changes in the relative prices of agricultural crops that are associated with
economies of scale and variation in climatic characteristics across states (that are static in
this short time period) would generate differences in the evolution of land concentration
across states.
To illustrate the differential effect of agricultural prices over time on the concentration of
landownership across states, consider the evolution of the price of cotton relative to the price
of corn over the period 18801940, as obtained from the NBER Macrohistory Database
(2007). The price of cotton relative to corn declined monotonically over the period of our
study. The price of a pound of cotton relative to a bushel of corn was 0321 in 1880, 0252 in
1900, 0236 in 1920, and 0155 in 1940, and indeed over this period, in regions that were
climatically more receptive to cotton production, the concentration of landownership held by
the largest farms declined. In particular, 29 states produced no cotton in 1860, and their
average change in land concentration was just 02% over the period 18801940. Among
states that produced some cotton in 1860, the average change in the land concentration of
the largest landowners was 26%. Cotton production was most prevalent in the South, with
this single crop accounting for over 40% of the total value of agricultural production in
45
Hence, the Hausman test indicates that the random effects specification is preferred. Furthermore, it is important to note that for
this sample of data, the random effects estimates are identical to the OLS results we report in column (4), due to the fact that the
estimated variance in
i
is 0.
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
Alabama, Arkansas, Florida, Georgia, Louisiana, Mississippi, South Carolina, and Texas.
Over this period the concentration of landownership by the largest farms declined in the
South from 20% in 1880 to 12% in 1900 and to only 8% in 1920.
Our instruments are therefore the interaction between state-specific, but time invariant,
climatic conditions and the nationwide changes in the price of cotton relative to corn.
46
The
climatic measures are derived from state data on temperature, rainfall, and a measure of
heating days (capturing the variability of temperature throughout the year) obtained from the
National Climatic Data Center (2007). As elaborated in Appendix D, using principal
components two distinct climatic measures are extracted from this data. The interaction of
these with the relative price data provides us with state-year-specific instruments for the
concentration of landownership.
These instruments appear to satisfy the exclusion restriction, since there is no evidence that
the human-capital intensity in the production of cotton over this period differs from the
average in all other agricultural crops, and changes in the relative price of cotton, therefore,
would not have a direct effect on education expenditure, but only indirectly through their
effect on concentration of landownership, and possibly via changes in income, that are
controlled for in the regressions.
The lagged differences of the instruments are used within a two-stage least squares
estimation to supply exogenous variation in the lagged difference of S
i,t1
. Included in our
specification are the lagged differences in log income per capita, the percentage of blacks,
the percentage of urban population, and period-specific dummies. Column (1) of Table 3
replicates the OLS results for comparison purposes, and column (2) reports the second-stage
results of the two-stage least squares estimation. As can be seen, the point estimate is now
larger in absolute value than in the OLS estimates, and remains significant at the 1% level.
The first stage results show that our instruments are quite strong in explaining variation in
S
i,t1
. As reported in the table, the F-test of the joint significance of the instruments has a
value of 1349, which is significant at less than 01%. Both instruments are individually
significant in the first stage at 1%. As we have two instruments, there is the possibility of
overidentification. However, a Sargan test of overidentification, as noted in the table, cannot
reject the null hypothesis that both instruments are uncorrelated with the error term
it
.
The results provide further support that we have identified a causal adverse effect of the
concentration of landownership on the provision of education across states during this
period of U.S. history.
7. CONCLUDING REMARKS
The proposed theory suggests that the concentration of landownership has been a major
hurdle in the emergence of human-capital promoting institutions and economic growth. The
rise in the demand for human capital in the process of industrialization and its effect on
human-capital formation and on the onset of the demographic transition have been the prime
forces in the transition from stagnation to growth. As the demand for human capital
emerged, differences in the concentration of landownership across countries generated
variations in the extensiveness of human-capital formation and therefore in the rapidity of
technological progress and the timing of the demographic transition, contributing to the
emergence of the great divergence in income per capita across countries. Land abundance,
which was beneficial in early stages of development, generated in later stages a hurdle for
46
The use of the price of cotton relative to wheat does not affect the results.
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
human-capital accumulation and economic growth among countries in which landownership
was unequally distributed, bringing about changes in the ranking of countries in the world
income distribution.
The central hypothesis of this research that inequality in the distribution of landownership
adversely affected the timing of education reform is examined and confirmed empirically,
utilizing variations in the distribution of landownership and educational expenditure across
states in the U.S. during the high-school movement. Furthermore, historical evidence
suggests that consistent with the proposed hypothesis, land reforms in Japan, Korea, Russia,
and Taiwan were associated with significant education reforms, and that variations in the
distribution of landownership across and within North and South America have been a
significant force in the emergence of sustained differences in human-capital formation and
economic growth.
The paper implies that differences in the evolution of social structures across countries may
reflect differences in the distribution of landownership. In particular, the dichotomy between
workers and capitalists is more likely to persist in land-abundant economies in which
landownership is unequally distributed. As argued by Galor and Moav (2006), due to the
complementarity between physical and human capital in production, the capitalists were
among the prime beneficiaries of the accumulation of human capital by the masses. They
had therefore the incentive to financially support public education that would sustain their
profit rates and would improve their economic well-being, although would ultimately
undermine their dynastys position in the social ladder and would bring about the demise of
the capitalistworkers class structure. As implied by the current research, the timing and the
extent of this social transformation depend on the economic interest of landowners. In
contrast to the Marxian hypothesis, this paper suggests that workers and capitalists are the
natural economic allies that share an interest in industrial development and therefore in the
implementation of growth enhancing human-capital promoting institutions, whereas
landowners are the prime hurdle for industrial development and social mobility.
Acknowledgments
We are grateful to two editors, four referees, and Yona Rubinstien for insightful and detailed comments. We thank
Daron Acemoglu, Josh Angrist, Andrew Foster, Eric Gould, Vernon Henderson, Saul Lach, Victor Lavy, Ross
Levine, Roger Myerson, Joel Mokyr, Daniele Paserman, Torsten Persson, Kenneth Sokolof, Nancy Stokey, David
Weil, Ivo Welch, Joseph Zeira, seminar participants at Brown, Cambridge, Clemson, CEFIR/NES, Chicago,
Cyprus, GREQAM, Oxford, Royal Holloway, Tel-Aviv, UCL, York, Warwick, and conference participants at the
NBER Summer Institute, 2003, the Economics of Water and Agriculture, 2003, Minervas conference, 2003,
ESSIM, Tarragona, 2004, CEPR Sardinia 2004, The Evolution of Institutions, 2006, Minerva-DEGIT Conference,
Jerusalem, 2006, CEPR, Annual Public Policy Symposium, Kiel, 2006, CIFAR, Tel Aviv, 2007, for helpful
comments. This research is supported by ISF grant 795/03. Galors research is supported by NSF Grant
SES-0004304.
References
ACEMOGLU D, ROBINSON JA. Economic Backwardness in Political Perspective. American
Political Science Review. 2006; 100:115131.
ACEMOGLU D, AGHION P, ZILIBOTTI F. Distance to Frontier, Selection, and Economic Growth.
Journal of the European Economic Association. 2006; 4:3774.
ACEMOGLU, D.; JOHNSON, S.; ROBINSON, JA. Institutions as the Fundamental Cause of Long-
Run Growth. In: AGHION, P.; DURLAUF, S., editors. Handbook of Economic Growth.
Amsterdam: Elsevier; 2005. p. 385-472.
ALESINA A, RODRIK D. Distributive Politics and Economic Growth. Quarterly Journal of
Economics. 1994; 109:465490.
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
ALESINA A, DEVLEESCHAUWER A, EASTERLY W, KURLAT S, WACZIARG R.
Fractionalization. Journal of Economic Growth. 2003; 8:155194.
ALTONJI JG, HAYASHI F, KOTLIKOFF LJ. Parental Altruism and Inter Vivos Transfers: Theory
and Evidence. Journal of Political Economy. 1997; 105:11211166.
BANERJEE A, IYER L. History Institutions and Economic Performance: The Legacy of Colonial
Land Tenure Systems in India. American Economic Review. 2005; 95:11901213.
BENABOU R. Inequality and Growth. NBER Macroeconomics Annual. 1996; 11:1174.
BENABOU R. Unequal Societies: Income Distribution and the Social Contract. American Economic
Review. 2000; 90:96129.
BENABOU R. Tax and Education Policy in a Heterogeneous-Agent Economy: What Levels of
Redistribution Maximize Growth and Efficiency? Econometrica. 2002; 70:481517.
BERTOCCHI G. The Law of Primogeniture and the Transition from Landed Aristocracy to Industrial
Democracy. Journal of Economic Growth. 2006; 11:4370.
BERTOCCHI G, CANOVA F. Did Colonization Matter for Growth? An Empirical Exploration into
the Historical Causes of Africas Underdevelopment. European Economic Review. 2002;
46:18511871.
BESLEY T, BURGESS R. Land Reform, Poverty Reduction, and Growth: Evidence from India.
Quarterly Journal of Economics. 2000; 115:389430.
BILKE NS, GORDON RJ. The Estimation of Prewar Gross National Product: Methodology and New
Evidence. Journal of Political Economy. 1989; 97:3892.
BONILLA, F. Brazil. In: Coleman, JS., editor. Education and Political Development. Princeton, NJ:
Princeton University Press; 1965.
BOURGUIGNON F, VERDIER T. Oligarchy, Democracy, Inequality and Growth. Journal of
Development Economics. 2000; 62:285313.
BOWLES S. Capitalist Development and Educational Structure. World Development. 1978; 6:783
796.
BREZIS E, KRUGMAN P, TSIDDON D. Leapfrogging: A Theory of Cycles in National
Technological Leadership. American Economic Review. 1993; 83:12111219.
CASELLI F, COLEMAN WJ. The U.S. Structural Transformation and Regional Convergence: A
Reinterpretation. Journal of Political Economy. 2001; 109:584616.
CHANDA A, DALGAARD CJ. Dual Economies and International Total Factor Productivity
Differences. Economica. 2008 Forthcoming.
CHENG, C. Land Reform in Taiwan. Taiwan: China Publishing Company; 1961.
COATSWORTH, JH. Notes on the Comparative Economic History of Latin America and the United
States. In: Bernecker, WL.; Tobler, HW., editors. Development and Underdevelopment in
America: Contrasts of Economic Growth in North and Latin America in Historical Perspective.
Berlin: de Gruyter; 1993. p. 10-30.
DEININGER K, SQUIRE L. New Ways of Looking at Old Issues: Inequality and Growth. Journal of
Development Economics. 1998; 57:259287.
DENNIS, G. Education. In: Florinsky, MT., editor. Encyclopedia of Russia and the Soviet Union. New
York: McGraw-Hill; 1961. p. 148-153.
DIAMOND, J. Guns, Germs, and Steel: The Fates of Human Societies. New York: Norton; 1997.
DOEPKE M. Accounting for Fertility Decline During the Transition to Growth. Journal of Economic
Growth. 2004; 9:347383.
DOEPKE M, ZILIBOTTI F. The Macroeconomics of Child Labor Regulation. American Economic
Review. 2005; 95:14921524.
EASTERLIN, RA. Regional Growth of Income: Long Term Tendencies. In: Kuznets, S.; Miller, AR.;
Easterlin, RA., editors. Population Redistribution and Economic Growth: United States. Vol. 2.
Philadelphia: American Philosophical Society; 1957. p. 1870-1950.p. 635-702.Analysis of
Economic Change
EASTERLY W. Inequality Does Cause Underdevelopment: Insights from a New Instrument. Journal
of Development Economics. 2007; 84:755776.
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
EASTERLY W, LEVINE R. Africas Growth Tragedy: Policies and Ethnic Divisions. Quarterly
Journal of Economics. 1997; 111:12031250.
EASTERLY W, LEVINE R. Tropics, Germs, and Crops: The Role of Endowments in Economic
Development. Journal of Monetary Economics. 2003; 50:339.
ECKERT, CJ. Korea Old and New: A History. Seoul: Ilchokak; 1990.
ENGERMAN S, SOKOLOFF KL. Institutions, Factor Endowments, and Paths of Development in the
New World. Journal of Economic Perspectives. 2000; 14:217232.
FERNANDEZ R, ROGERSON R. Income Distribution, Communities, and the Quality of Public
Education. Quarterly Journal of Economics. 1996; 111:135164.
FLORINSKY, MT. Encyclopedia of Russia and the Soviet Union. New York: McGraw-Hill; 1961.
FOSTER AD, ROSENZWEIG MR. Technical Change and Human-Capital Returns and Investments:
Evidence from the Green Revolution. American Economic Review. 1996; 86:931953.
GALOR, O. Unified Growth Theory: From Stagnation to Growth. In: Aghion, P.; Durlauf, S., editors.
Handbook of Economic Growth. Amsterdam: Elsevier; 2005. p. 171-294.
GALOR O, MOAV O. Natural Selection and the Origin of Economic Growth. Quarterly Journal of
Economics. 2002; 117:11331192.
GALOR O, MOAV O. From Physical to Human Capital: Inequality in the Process of Development.
Review of Economic Studies. 2004; 71:10011026.
GALOR O, MOAV O. Das Human-Kapital: A Theory of the Demise of Class Structure. Review of
Economic Studies. 2006; 73:85117.
GALOR O, MOUNTFORD A. Trade and the Great Divergence: The Family Connection. American
Economic Review. 2006; 96:299303.
GALOR O, MOUNTFORD A. Trading Population for Productivity: Theory and Evidence. Review of
Economic Studies. 2008 Forthcoming.
GALOR O, WEIL DN. Population, Technology and Growth: From the Malthusian Regime to the
Demographic Transition. American Economic Review. 2000; 90:806828.
GALOR O, ZEIRA J. Income Distribution and Macroeconomics. Review of Economic Studies. 1993;
60:3552.
GERBER J. Public-School Expenditures in the Plantation States, 1910. Explorations in Economic
History. 1991; 28:309322.
GHOSAL, S.; PROTO, E. Democracy, Intra-Elite Conflict and Risk Sharing. Mimeo: Warwick; 2005.
Why did (not) the East Extend the Franchise?.
GLAESER EL, SHLEIFER A. Legal Origins. Quarterly Journal of Economics. 2002; 117:11921230.
GLAESER EL, LA PORTA R, LOPEZ-DE-SILANES F, SHLEIFER A. Do Institutions Cause
Growth? Journal of Economic Growth. 2004; 9:271303.
GOLDIN C. Americas Graduation from High School: The Evolution and Spread of Secondary
Schooling in the Twentieth Century. Journal of Economic History. 1998; 58:345374.
GOLDIN C. Egalitarianism and the Returns to Eduction during the Great Transformation of American
Education. Journal of Political Economy. 1999; 107:S65S94.
GOLDIN, C.; KATZ, LF. Working Paper No 6144. NBER; Cambridge, MA: 1997. Why the United
States Led in Education: Lessons from Secondary School Expansion, 1910 to 1940.
GRADSTEIN M. Inequality, Democracy, and the Protection of Property Rights. Economic Journal.
2007; 117:252269.
GUBBINS, JH. The Making of Modern Japan. Wilmington, DE: Scholarly Resources; 1973.
HALL RE, JONES CI. Why Do Some Countries Produce So Much More Output Per Worker Than
Others? The Quarterly Journal of Economics. 1999; 114:83116.
HAYAMI, Y. A Century of Agricultural Growth in Japan. Minneapolis: University of Minnesota
Press; 1975.
HIBBS DE, OLSSON O. Biogeography and Long-Run Economic Development. European Economic
Review. 2005; 49:909938.
HOWITT P, MAYER-FOULKES D. R&D, Implementation and Stagnation: A Schumpeterian Theory
of Convergence Clubs. Journal of Money Credit and Banking. 2005; 37:147177.
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
JOHNSON, WHE. Russias Educational Heritage. New York: Octagon Books; 1969.
JONES, EL. Economies and Geopolitics in the History of Europe and Asia. Cambridge: Cambridge
University Press; 1981. The European Miracle: Environments.
LAGERLOF, NP. Slavery. Mimeo: York University; 2003.
LAGERLOF NP. The Galor-Weil Model Revisited: A Quantitative Exercise. Review of Economic
Dynamics. 2006; 9:116142.
LANDES, DS. The Wealth and Poverty of Nations. New York: Norton; 1998.
LEBOT, I. Elementos Papa La Historia de la Educacion en Colombia en el Siglo XX. Bogota: Talleres
de DANE; 1972.
LIN, CJ. The Republic of China (Taiwan). In: Thomas, RM.; Postlethwaite, TN., editors. Schooling in
East Asia. Oxford: Pergamon Press; 1983. p. 101-136.
LINDBECK A, WEIBULL J. Balanced-Budget Redistribution as the Outcome of Political
Competition. Public Choice. 1987; 52:273297.
LINDERT PH. Unequal English Wealth Since 1670. Journal of Political Economy. 1986; 94:1127
1162.
LIZZERI A, PERSICO N. Why Did the Elites Extend the Suffrage? Democracy and the Scope of
Government, With an Application to Britains Age of Reform. Quarterly Journal of Economics.
2004; 119:707765.
LLAVADOR H, OXOBY RJ. Partisan Competition, Growth, and the Franchise. Quarterly Journal of
Economics. 2005; 120:11551192.
MADDISON, A. The World Economy: A Millennia Perspective. Paris: OECD; 2001.
MANUELLI, RE.; SESHADRI, A. Human Capital and the Wealth of Nations. Mimeo: Wisconsin;
2005.
MARGO, RA. 18801950: An Economic History. Chicago: University of Chicago Press; 1990. Race
and Schooling in the South.
MOKYR, J. The Lever of Riches: Technological Creativity and Economic Progress. New York:
Oxford University Press; 1990.
MOKYR, J. The Gifts of Athena: Historical Origins of the Knowledge Economy. Princeton: Princeton
University Press; 2002.
National Bureau of Economic Research. Macrohistory Database. 2007. http://www.nber.org/databases/
macrohistory/contents/
NORTH, DC. Structure and Change in Economic History. New York: W.W. Norton & Co; 1981.
NUGENT, JB.; ROBINSON, JA. Are Endowments Fate?. CEPR; London: 2002. Discussion Paper
No. 3206
OLSON, M. The Rise and Decline of Nations: Economic Growth, Stagflation, and Social Rigidities.
New Haven: Yale University Press; 1982.
PARENTE, S.; PRESCOTT, EC. Barriers to Riches. Cambridge, MA: MIT Press; 2000.
PASSIN, H. Japan. In: Coleman, JS., editor. Education and Political Development. Princeton, NJ:
Princeton University Press; 1965. p. 272-312.
PEROTTI R. Growth, Income Distribution, and Democracy: What the Data Say. Journal of Economic
Growth. 1996; 1:149187.
PERSSON T, TABELLINI G. Is Inequality Harmful for Growth? American Economic Review. 1994;
84:600621.
PERSSON, T.; TABELLINI, G. Political Economics: Explaining Economic Policy. Cambridge, MA:
MIT Press; 2000.
RODRIK D, SUBRAMANIAN A, TREBBI F. Institutions Rule: The Primacy of Institutions Over
Geography and Integration in Economic Development. Journal of Economic Growth. 2004;
9:131165.
SACHS, J.; WARNER, AM. Working Paper No 5398. NBER; Cambridge, MA: 1995. Natural
Resource Abundance and Economic Growth.
SAH-MYUNG, H. The Republic of Korea (South Korea). In: Thomas, RM.; Postlethwaite, TN.,
editors. Schooling in East Asia. Oxford: Pergamon Press; 1983. p. 205-240.
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
U.S. Bureau of the Census. Historical Statistics of the United States: Colonial Times to 1970. White
Plains, NY: Kraus International Publications; 1989.
U.S. Bureau of Education. Report of the Commissioner of Education. Washington, DC: Government
Printing Office; 18701950.
U.S. Department of Agriculture. Profile of Hired Farm Workers. Annual Averages. 1998
U.S. Department of Commerce National Climatic Data Center. US Climate Normals. U.S. Department
of Agriculture; Washington, DC: 2007. available online at http://cdo.ncdc.noaa.gov/cgi-bin/
climatenormals.pl
VAUGHAN, MK. The State, Education, and Social Class in Mexico, 18801928. DeKalb, IL:
Northern Illinois University Press; 1982.
VOIGTLANDER N, VOTH HJ. Why England? Demographic Factors, Structural Change and Physical
Capital Accumulation During the Industrial Revolution. Journal of Economic Growth. 2006;
11:319361.
WRIGHT, G. Old South New South: Revolutions in the Southern Economy Since the Civil War. New
York: Basic Books; 1986.
YOONG-DEOK J, KIM YY. Land Reform, Income Redistribution, and Agricultural Production in
Korea. Economic Development and Cultural Change. 2000; 48:253268.
APPENDIX A. LANDOWNERS PREFERRED TAX RATE: COBBDOUGLAS
AGRICULTURAL TECHNOLOGY
Lemma A3
The elasticity of
t
with respect to .
Proof
Suppose not. Suppose that . Since a rise in and a decline in
t
imply a rise in w and a reduction in the optimal number of workers in agriculture and hence
a rise in
t
. A contradiction. Suppose that . since a rise in and a
more than proportional rise in
t
implies a decline in w
t
and a rise in the optimal number of
workers in agriculture and hence a decline in
t
. A contradiction. ||
Proposition A6
If the agricultural production function is then landowners desirable tax
rate .
Proof
As follows from (3), noting that L
t
= 1
t
,
(A.1)
Hence, along the factor price frontier
(A.2)
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
Let . It follows from (18) and (20) that w
t
= (1 )
t
and
(A.3)
Since the wage paid to each worker is equal in the two sectors, it follows from (A.1) that
(A.4)
and hence
(A.5)
Note that
t
is determined endogenously such that
t
(0, 1).
Since landlords income in period t is , it follows that the
aggregate income of landowners, , is
(A.6)
where w
t
= (1 a)
t
, is the wage, is the share of capital in the industrial output,
and is the share of capital owned by landowners. Substituting (A.3) and (A.5) into (A.6)
(A.7)
Hence,
(A.8)
Hence,
If, however, , replacing for in (A.7),
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
which is strictly increasing in in
t
.
Hence, if landlords income is strictly increasing in
t
and it follows from Lemma
A3 that landowners prefer the highest possible value for , and therefore . Noting
that and it follows that for .
If, however, landowners income is a convex function of
t
, implying they prefer
either the maximal or the minimal value of
t
. Therefore, it follows from Lemma A3 that
landowners prefer the highest or lowest possible value for , and hence . ||
APPENDIX B. CONSTRUCTION OF THE LAND SHARE DISTRIBUTION
VARIABLES
The Census reports the distribution of number of farms by bin size (e.g. less than 20 acres,
2049 acres, 5099 acres, 100499 acres, 500999 acres, and greater than 1000 acres).
Since the Censuses do not report identical bin sizes for 1880, 1900, and 1920, we aggregate
farms into the previously listed six categories that are comparable across years.
To calculate S
it
, we require information on the actual size of farms within each size category
reported by the Census, but this is not directly available in 1880 or 1900. We therefore
assume that each farm within a given category is of the average size of that category. For
example, each farm in the range 2049 acres is assumed to be 345 acres. For the category of
farms greater than 1000 acres, we assume that each farm is exactly 1000 acres. The results
are not sensitive to alternate choices. While data on farm size within categories is available
for 1920, it does not differ greatly from the assumed average values, and the results are not
sensitive to the use of the actual farm size data for 1920.
Initially we calculate a Gini coefficient of farm size distribution. There are N size categories,
numbered from 1 to 6 in order of increasing size of farms. Let f
i
be the share of all farms
that are in category i. Let a
i
be the share of all acreage that is in category i. Let ,
denotes the share of farms that are of size i or smaller. Similarly, . By definition,
F
6
= A
6
= 1. It can be shown that the Gini coefficient, G, can be calculated as follows:
To perform our calculations of S
it
, the land share of the largest landowners, we use a simple
parameterization of the Lorenz curve. This is denoted where s
L
is the share of land
and s
F
is the share of farms. The parameter is related to the Gini coefficient in the
following manner, = (1 + G)/(1 G). Thus given the Gini coefficient we can derive the
parameter for each state. Given , we calculate the minimum number of farms in 1880 that
constitute 20% of total farm land as TopFarms
1880
= Farms
1880
(1 (080)
1/
) where
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
Farms
1880
is the total number of farms in 1880, is the Lorenz parameter from 1880, and
TopFarms
1880
is the number of large farms constituting 20% of all land. Note that we can
utilize other choice of percentage (i.e. 5%, 10% etc.) in place of 20%.
We can then track how the share of land held by the largest number of farms evolves over
time, where the number of these farms is held constant at TopFarms
1880
. By construction,
this share of land is 20% in 1880. For subsequent years, the share can be calculated as S
it
= 1
(1 TopFarms
i,1880
/Farms
it
)
it
where Farms
it
measures the total number of farms in year
t, and
it
measures the coefficient on the Lorenz curve from year t in state i. One advantage
of this calculation is that it is independent of the average farm size between states, which
varies incredibly across the U.S. based on geographic conditions rather than differences in
inequality.
The current method allows for a smooth distribution of farm sizes over the whole range of
farms. However, we could alternately calculate the S
it
variable by going directly to the size
distribution data in the U.S. Census. This would assume that each farm within each size
category of farm is of an identical size. Starting from the largest bin size (greater than 1000
acres) in 1880 and working down the bin sizes if necessary, we count how many farms
account for 20% of farm land. We can then take this number of top farms in subsequent
years and ask how much land is accounted for by this same number of farms, again working
from the top of the distribution down. One disadvantage of this method is that it depends on
the assumed average size of farms over 1000 acres, which is not reported by the U.S.
Census. The parameterization method we utilize is less sensitive to this lack of information.
APPENDIX C. DATA SOURCES
Education Expenditures
This is obtained from the Historical Statistics of the U.S. for 1920 and 1940, and from the
U.S. Bureau of Education, Report of the Commissioner of Education for 1900. These
expenditures were converted to 1920 dollars using the GDP deflator from Bilke and Gordon
(1989).
Expenditure per Child
The number of children in a year is taken from the U.S. Census. Consistent across all states,
the available age ranges are 520 years in 1900 and 720 years in 1920.
Income per Capita
These are estimates Richard Easterlin (1957), Population Redistribution, and Economic
Growth: U.S. 18701950, edited by Kuznets and Thomas. Details of the construction can be
obtained from this source. Income per capita is converted to 1920 dollars using the GDP
deflator from Bilke and Gordon (1989).
Percentage of Blacks
This is taken from the U.S. Census for the relevant years.
Urban Percentage
From the U.S. Census for the relevant years.
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
Regions
The Northeast is defined as Connecticut, Massachusetts, New Hampshire, New Jersey, New
York, Maine, Pennsylvania, Rhode Island, and Vermont. The Midwest is Illinois, Indiana,
Iowa, Kansas, Michigan, Minnesota, Missouri, Nebraska, Ohio, and Wisconsin. The South
is Alabama, Arkansas, Delaware, Florida, Georgia, Kentucky, Louisiana, Maryland,
Mississippi, North Carolina, South Carolina, Tennessee, Texas, Virginia, and West Virginia.
The West includes California, Montana, Oregon, Utah, Washington, and Wyoming.
Excluded States
States that are not included in the sample due to data limitations are: Alaska, Arizona,
Colorado, Hawaii, Idaho, Nevada, New Mexico, North Dakota, South Dakota, and
Oklahoma.
APPENDIX D. CONSTRUCTION OF INSTRUMENTS
The instrument used to identify the effect of land distribution on education is composed of a
geographic element and a relative price element. The geographic element is derived from
state level data on temperature, rainfall, and heating requirements obtained from the U.S.
Department of Commerce National Climatic Data Center (NCDC) (2007), specifically the
U.S. Climate Normals. The Climate Normals provide annual average temperature and
rainfall for individual climatic regions within states. Annual heating requirements are
derived from the average pattern of temperatures throughout the year and capture the
intensity of temperature differences within the year (as opposed to the simple average). The
heating requirements are reported by regions within states as well. For each measure, the
value reported is the average value over the years 1931 through 2000.
To calculate state level values, the regional data are weighted within each state by the area
of the region as reported by he NCDC. Examining the data, the three measures of climate
are significantly correlated with each other. The correlation of temperature with heating days
is 099, temperature with rainfall is 049, and rainfall with heating days is 048. All
correlations are significant at less than 1%. Using these three collinear terms together may
falsely inflate the significance of the first stage regressions, so to extract the most
explanatory power from the climate data while limiting the number of variables, a principal
components analysis is used. The first component accounts for 78% of the variation
observed in the three variables, with an eigenvalue of 235. The loadings on the first
component are 0628 for heating days, 0630 for temperature, and 0457 for annual rainfall.
The second component accounts for 21% of total variation, with an eigenvalue of 064. The
loadings are 0327 for heating days, 0319 for temperature, and 0890 for annual rainfall.
Combined, the two principal components capture 99% of the variation in climatic data, with
the first component picking up mainly variation associated with temperature and the second
component picking up mainly rainfall.
The two climatic components are state specific, and time invariant. To generate a time-
varying instrument, we have combined the climate data with information on the relative
price of cotton. As explained in the text, the concept is that increases in cotton prices would
induce land concentration, but only in those places, which were geographically suited to
cotton production in the first place. The price data comes from the National Bureau of
Economic Research (NBER) Macrohistory Database. Historical price levels for cotton are
obtained as an average of the monthly wholesale price from New York (NBER series
04006), in cents per pound. These prices are 1204 cents in 1880, 964 in 1900, 3390 in
1920, and 1038 in 1940. The price of corn, also an average of monthly wholesale prices
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
from Chicago (NBER series 04005), in dollars per bushel. The specific prices are 0375
dollars in 1880, 0383 dollars in 1900, 1439 in 1920, and 0670 in 1940.
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
Figure 1.
The evolution of income per capita before and after the implementation of education reforms
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
Figure 2.
Overtaking. Country A is relatively richer in land, however, due to land inequality it fails to
implement efficient schooling and is overtaken by country B
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
Figure 3.
Changes in education expenditure per child and changes in the concentration in
landownership
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
T
A
B
L
E

1
C
o
r
r
e
l
a
t
i
o
n
s

o
f

v
a
r
i
a
b
l
e
s

l
n
e
i
t
S
i
,
t

l
n

y
i
,
t
B
i
,
t
U
i
,
t

l
n
e
i
t
1
0
0
0
S
i
,
t
3
2
4
*
*
1
0
0
0

l
n

y
i
,
t
1
0
4
1
1
*
*
0
1
3
1
0
0
0
B
i
,
t
4
6
0
*
*
0
1
7
3
1
7
3
1
0
0
0
U
i
,
t
0
3
4
0
3
2
2
*
*
0
1
1
2
2
5
0
*
*
1
0
0
0
*
*
i
n
d
i
c
a
t
e
s

s
i
g
n
i
f
i
c
a
n
c
e

a
t

t
h
e

5
%

l
e
v
e
l

a
n
d
*
a
t

t
h
e

1
0
%

l
e
v
e
l
.
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
T
A
B
L
E

2
S
p
e
c
i
f
i
c
a
t
i
o
n
s

f
o
r

c
h
a
n
g
e
s

i
n

p
e
r
-
c
h
i
l
d

e
d
u
c
a
t
i
o
n

e
x
p
e
n
d
i
t
u
r
e
E
x
p
l
a
n
a
t
o
r
y

v
a
r
i
a
b
l
e
s
D
e
p
e
n
d
e
n
t

v
a
r
i
a
b
l
e
:

c
h
a
n
g
e

i
n

l
o
g

e
d
u
c
a
t
i
o
n
a
l

e
x
p
e
n
d
i
t
u
r
e

p
e
r

c
h
i
l
d

(

l
n
e
i
t
)
O
L
S

(
1
)
O
L
S

(
2
)
O
L
S

(
3
)
O
L
S

(
4
)
O
L
S

(
5
)
O
L
S

(
6
)
C
h
a
n
g
e

i
n

l
a
n
d

c
o
n
c
e
n
t
r
a
t
i
o
n

(
S
i
,
t
1
)
7
1
*
*
*

(
0
9
9
)
6
7
*
*
*

(
0
8
6
)
1
6
*
*
*

(
0
7
5
)
1
2
*
*
*

(
0
7
8
)
3
4
*
*
*

(
0
8
0
)
6
8
*

(
2
1
7
)
C
h
a
n
g
e

i
n

i
n
c
o
m
e

p
e
r

c
a
p
i
t
a

(

l
n

y
i
,
t
1
)
0
8
4
*
*
*

(
0
1
5
)
0
7
2
*
*
*

(
0
1
3
)
0
7
2
*
*
*

(
0
1
3
)
0
7
2
*
*
*

(
0
1
7
)
0
7
1
*

(
0
4
1
)
C
h
a
n
g
e

i
n

%

o
f

t
h
e

b
l
a
c
k

p
o
p
u
l
a
t
i
o
n

(
B
i
,
t
1
)
7
4
*
*
*

(
0
5
9
)
7
8
*
*
*

(
0
7
3
)
9
0
*
*
*

(
0
9
6
)
1
3
*
*

(
2
1
7
)
C
h
a
n
g
e

i
n

%

o
f

t
h
e

u
r
b
a
n

p
o
p
u
l
a
t
i
o
n

(
U

i
,
t
1
)
0
5

(
0
3
2
)
6
6
*

(
0
4
0
)
1
2

(
0
6
9
)
N
a
t
i
o
n
a
l

t
i
m
e

f
i
x
e
d

e
f
f
e
c
t
s
N
o
N
o
N
o
N
o
Y
e
s
N
o
S
t
a
t
e

f
i
x
e
d

e
f
f
e
c
t
s

(
l
i
n
e
a
r

t
i
m
e

t
r
e
n
d
)
N
o
N
o
N
o
N
o
N
o
Y
e
s
R
2
0
1
1
0
2
7
0
3
9
0
3
9
0
4
8
0
3
8
H
a
u
s
m
a
n

s
t
a
t
i
s
t
i
c
2
1
6
H
a
u
s
m
a
n

p
-
v
a
l
u
e
0
7
1
N
o
t
e
s
:
*
*
*
d
e
n
o
t
e
s

s
i
g
n
i
f
i
c
a
n
c
e

a
t

t
h
e

1
%

l
e
v
e
l
,
*
*
a
t

5
%
,

a
n
d
*
a
t

1
0
%
.
S
.
E
.

a
r
e

a
d
j
u
s
t
e
d

f
o
r

c
l
u
s
t
e
r
i
n
g

a
t

t
h
e

s
t
a
t
e

l
e
v
e
l
.

A
l
l

r
e
g
r
e
s
s
i
o
n
s

h
a
v
e

a

t
o
t
a
l

o
f

7
9

o
b
s
e
r
v
a
t
i
o
n
s

f
r
o
m

4
1

s
t
a
t
e
s
.

T
h
e

n
u
l
l

h
y
p
o
t
h
e
s
i
s

o
f

t
h
e

H
a
u
s
m
a
n

t
e
s
t

i
n

c
o
l
u
m
n

(
6
)

i
s

t
h
a
t

t
h
e
r
e

i
s

n
o

s
y
s
t
e
m
a
t
i
c

d
i
f
f
e
r
e
n
c
e
i
n

t
h
e

p
o
i
n
t

e
s
t
i
m
a
t
e
s

b
e
t
w
e
e
n

c
o
l
u
m
n
s

(
4
)

a
n
d

(
6
)
;

t
h
e

s
t
a
t
i
s
t
i
c

i
s

d
i
s
t
r
i
b
u
t
e
d

2
(
4
)
.

D
a
t
a

o
n

l
o
g

e
d
u
c
a
t
i
o
n

e
x
p
e
n
d
i
t
u
r
e
s

p
e
r

c
h
i
l
d

(
l
n
e
i
t
)

i
s

f
r
o
m

t
h
e

y
e
a
r
s

1
9
0
0
,

1
9
2
0
,

a
n
d

1
9
4
0
,

a
n
d

d
a
t
a

f
o
r

a
l
l

e
x
p
l
a
n
a
t
o
r
y
v
a
r
i
a
b
l
e
s

i
s

l
a
g
g
e
d
,

t
a
k
e
n

f
r
o
m

t
h
e

t
h
e

y
e
a
r
s

1
8
8
0
,

1
9
0
0
,

a
n
d

1
9
2
0
.
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
N
I
H
-
P
A

A
u
t
h
o
r

M
a
n
u
s
c
r
i
p
t
TABLE 3
Instrumental variable specifications for changes in per-child education expenditure
Explanatory variables Dependent variable: change in log educational expenditure per child ( lne
it
)
OLS (1) 2SLS (2)
Change in land concentration (S
i,t1
)
234
***
(080) 323
***
(091)
Change in income per capita ( ln y
i,t1
)
072
***
(017) 072
***
(017)
Change in % of the black population (B
i,t1
)
290
***
(096) 258
***
(092)
Change in % of the urban population (U
i,t1
)
066
*
(040)
051 (037)
National time fixed effects Yes Yes
R
2
048
First stage F-statistic 1349
First stage p-value <0001
Sargan test statistic 120
Sargan test p-value 027
Notes:
***
denotes significance at the 1% level,
**
at 5%, and
*
at 10%.
S.E.s are adjusted for clustering at the state level. All regressions have 79 total observations, from 41 states. Data on log education expenditures per
child (lne
it
) are from 1900, 1920, and 1940, and data for all control variables is from 1880, 1900, and 1920. In column (2), the instruments for the
difference in land concentration (S
i,t1
) are two variables that interact state-specific climate conditions with the price of cotton relative to the
price of corn.. The F-test in column (2) tests the null hypothesis that the coefficient on both instruments in the first stage is 0; the statistic is
distributed F(2, 72). The Sargan test has a null hypothesis that both instruments are uncorrelated with the error term; the statistic is distributed
2
(1).

Ni Hms 479919

Hochgeladen von

Dokumentinformationen

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Ni Hms 479919

Hochgeladen von

Copyright:

Verfügbare Formate

Inequality in Landownership, the Emergence of Human-Capital

Promoting Institutions, and the Great Divergence

be the first period in which the efficient tax policy, , is

(i.e. the later

Das könnte Ihnen auch gefallen